Set the chart's placeholder version to 0.41.0

Cosmetic: make helm-package passes --version and --app-version from the
tag, so these two fields decide nothing about what gets published. Still
done, as with ead5df1 (0.40.0) and 0f88574 (0.39.0) before it, because a
tree heading for v0.41.0 that still says 0.40.0 tells its reader
something false.

.gitea/workflows/ci.yaml

@@ -125,6 +125,9 @@ jobs:
       - name: Secret scan (gitleaks)
         run: make security-secrets
 
+      - name: Code security scan (gosec)
+        run: make security-code
+
   # Host mode, no `container:`: helm is baked into the runner image, and a container job
   # could not install it -- get.helm.sh is unreachable from the dind bridge. Same reason
   # release.yaml's chart job runs on the host.

Dockerfile

@@ -24,4 +24,10 @@ FROM scratch
 COPY --from=builder /etc/ssl/certs/ca-certificates.crt /etc/ssl/certs/ca-certificates.crt
 COPY --from=builder /terdut /terdut
 EXPOSE 8080
+# Numeric, not a name: scratch has no /etc/passwd for one to resolve against,
+# and Docker's USER accepts a bare UID:GID without it. 65532 is the common
+# "nonroot" convention (distroless's own uid), chosen so the chart's pod
+# securityContext (runAsNonRoot, runAsUser: 65532) matches what the image
+# already runs as rather than fighting it.
+USER 65532:65532
 ENTRYPOINT ["/terdut"]

Makefile

@@ -143,6 +143,7 @@ BUILDX_BUILDER ?= terdut
 TRIVY_VERSION       := 0.73.0
 GOVULNCHECK_VERSION := v1.1.4
 GITLEAKS_VERSION    := v8.30.0
+GOSEC_VERSION       := v2.29.0
 
 # --pull, not --no-cache: refresh the base image without discarding the layer cache.
 DOCKER_BUILD_FLAGS ?= --pull
@@ -222,6 +223,26 @@ release: push helm-package helm-push ## Publish image + chart (the workflow's on
 security-go: ## Scan Go deps for known CVEs (govulncheck)
 	go run golang.org/x/vuln/cmd/govulncheck@$(GOVULNCHECK_VERSION) ./...
 
+# Code-level, not dependency- or secret-level: gosec reads this repo's own source for
+# known-dangerous patterns (weak crypto, SQL/command injection shapes, insecure file
+# permissions, …) rather than its module graph or working tree for leaked credentials,
+# which is what security-go and security-secrets above already cover.
+#
+# G104 (unchecked error) is excluded. Every hit it found here on first run was this
+# codebase's existing, deliberate idiom for a best-effort write or an already-reviewed
+# json.Unmarshal of this server's own JSONB (see the "best-effort" comments in
+# middleware.go and the //nolint:errcheck lines in alerts.go/deadman.go) -- a style that
+# predates gosec and that G104 cannot distinguish from a mistake. Reaching the same
+# green result by adding a dozens of individual #nosec comments would not add
+# information; it would just make a future *real* G104 regression one more suppressed
+# line instead of a visible one. Same reasoning as the chi-advisories note on
+# security-go above: what gosec reports here (nothing, beyond G104) is the useful
+# property, not a loophole. -exclude-generated skips web.go's embedded, build-time-only
+# assets.
+.PHONY: security-code
+security-code: ## Scan this repo's own source for risky patterns (gosec)
+	go run github.com/securego/gosec/v2/cmd/gosec@$(GOSEC_VERSION) -exclude-generated -exclude=G104 ./...
+
 # --no-git scans the working tree rather than the history, so this catches a secret on the
 # way in. It is not a history audit and finding nothing here says nothing about what is
 # already committed. --redact because the finding is printed into a CI log.

README.md

@@ -984,9 +984,11 @@ name: degrade unknown values to "resolved, reason unknown".
 | `created_at` | timestamp | |
 
 Types written today: `triggered`, `alert_added`, `alert_resolved`,
-`acknowledged`, `unacknowledged`, `assigned`, `snoozed`, `unsnoozed`, `resolved`,
-`note`, `notified`, `notify_failed`, `deadman_silent`. On an `assigned` event
-`user_id` is the **assignee**, not the actor. New types may be added; render
+`acknowledged`, `unacknowledged`, `assigned`, `archived`, `unarchived`, `snoozed`,
+`unsnoozed`, `resolved`, `note`, `notified`, `notify_failed`, `deadman_silent`. On an
+`assigned` event `user_id` is the **assignee**, not the actor; the actor is in
+`actor_user_id`/`actor_username` or `actor_service_account_id`/`actor_service_account_name`
+(absent on assignments made before they were recorded). New types may be added; render
 unknown ones generically rather than dropping them.
 
 On `notified` and `notify_failed`, `detail` carries the notification kind

charts/terdut-server/Chart.yaml

@@ -15,5 +15,5 @@ type: application
 # appVersion and image.tag in values.yaml no longer agree, and that is not an oversight:
 # image.tag stays "latest", which is what a local install actually pulls. appVersion is
 # metadata and drives nothing.
-version: 0.35.1
-appVersion: "v0.35.1"
+version: 0.41.0
+appVersion: "v0.41.0"

charts/terdut-server/templates/deployment.yaml

@@ -6,26 +6,49 @@ metadata:
   labels:
     {{- include "terdut-server.labels" . | nindent 4 }}
 spec:
-  replicas: 1
+  replicas: {{ .Values.replicaCount }}
   selector:
     matchLabels:
       {{- include "terdut-server.selectorLabels" . | nindent 6 }}
-  # Recreate, not RollingUpdate, even though the PVC that forced it is gone: the
-  # sweeper and the notifier are unsynchronised singletons, and two replicas
-  # overlapping during a rollout would both page for the same incident.
+  # RollingUpdate, not Recreate: the sweeper, notifier and migration runner
+  # each take a Postgres advisory lock around their own pass, and new-incident
+  # creation on the first webhook for a brand-new groupKey resolves its own
+  # insert conflict -- so two replicas overlapping during a rollout no longer
+  # double-page, race a migration, or drop a webhook payload (v0.36.0). No
+  # explicit maxUnavailable/maxSurge: the 25%/25% default rounds to 0/1 at
+  # replicaCount: 2, which is zero-downtime already.
   strategy:
-    type: Recreate
+    type: RollingUpdate
   template:
     metadata:
       labels:
         {{- include "terdut-server.selectorLabels" . | nindent 8 }}
     spec:
       enableServiceLinks: false
+      # Pod-wide default; both containers below run as this UID regardless of
+      # what their own image would otherwise pick (postgres:17-alpine's
+      # pg_isready needs no particular user, and 65532 is what the app image
+      # itself runs as now — see the Dockerfile's USER). seccompProfile here
+      # rather than per-container: there is no reason it would ever differ
+      # between them.
+      securityContext:
+        runAsNonRoot: true
+        runAsUser: 65532
+        runAsGroup: 65532
+        seccompProfile:
+          type: RuntimeDefault
       {{- if .Values.database.waitForPostgres.enabled }}
       initContainers:
         - name: wait-for-postgres
           image: "{{ .Values.database.waitForPostgres.image.repository }}:{{ .Values.database.waitForPostgres.image.tag }}"
           imagePullPolicy: {{ .Values.database.waitForPostgres.image.pullPolicy }}
+          # No capability this loop needs, and nothing in it writes to disk:
+          # sh, pg_isready, echo and sleep all run read-only.
+          securityContext:
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            capabilities:
+              drop: ["ALL"]
           env:
             - name: TERDUT_DB_DSN
               value: {{ required "database.dsn is required" .Values.database.dsn | quote }}
@@ -42,6 +65,13 @@ spec:
         - name: terdut-server
           image: "{{ .Values.image.repository }}:{{ .Values.image.tag }}"
           imagePullPolicy: {{ .Values.image.pullPolicy }}
+          # scratch, nothing to write: the binary keeps no local state and
+          # writes nothing to disk, so the root filesystem can stay read-only.
+          securityContext:
+            allowPrivilegeEscalation: false
+            readOnlyRootFilesystem: true
+            capabilities:
+              drop: ["ALL"]
           ports:
             - name: http
               containerPort: {{ .Values.service.port }}

charts/terdut-server/values.yaml

@@ -1,3 +1,10 @@
+# Safe above 1 since v0.36.0: the sweeper, notifier and migration runner each
+# take a Postgres advisory lock around their own pass, and a webhook that
+# loses the race to open a brand-new incident attaches to the winner's row
+# instead of dropping its payload. An image older than v0.36.0 does not have
+# these guards -- do not raise this against one.
+replicaCount: 2
+
 networking:
   hostname: "terdut.example.com"
   servicePort: 8080

internal/api/advisory_lock_test.go

@@ -0,0 +1,122 @@
+package api
+
+// This file is internal (package api, not api_test) because withAdvisoryLock is
+// unexported and these tests exercise its locking semantics directly rather than
+// through the full StartArchiver/StartNotifier loop, which would make the "does
+// not run while held" case timing-dependent instead of deterministic. It opens a
+// plain connection to TERDUT_TEST_DSN rather than reusing testdb_test.go's
+// newTestDB, since that helper lives in the separate, already-compiled
+// api_test package and a Postgres advisory lock needs no schema or migration
+// to exercise.
+
+import (
+	"context"
+	"database/sql"
+	"os"
+	"testing"
+
+	_ "github.com/jackc/pgx/v5/stdlib"
+)
+
+// advisoryTestDB opens a plain, unmigrated connection to the test database. An
+// unset DSN fails rather than skips, matching testdb_test.go's rationale: a
+// suite that quietly tests nothing is worse than one that does not run.
+func advisoryTestDB(t *testing.T) *sql.DB {
+	t.Helper()
+
+	dsn := os.Getenv("TERDUT_TEST_DSN")
+	if dsn == "" {
+		t.Fatalf("TERDUT_TEST_DSN is not set: these tests need Postgres.\n" +
+			"Run `make test-db` for a local one, then\n" +
+			"  export TERDUT_TEST_DSN=postgres://terdut:terdut@localhost:5432/terdut_test?sslmode=disable")
+	}
+
+	db, err := sql.Open("pgx", dsn)
+	if err != nil {
+		t.Fatalf("connect to TERDUT_TEST_DSN: %v", err)
+	}
+	t.Cleanup(func() { db.Close() })
+	return db
+}
+
+func TestWithAdvisoryLock_RunsWhenFree(t *testing.T) {
+	db := advisoryTestDB(t)
+	ctx := context.Background()
+
+	ran := false
+	withAdvisoryLock(ctx, db, archiverLockKey, "test", func() { ran = true })
+
+	if !ran {
+		t.Fatal("fn did not run although the lock was free")
+	}
+}
+
+func TestWithAdvisoryLock_SkipsWhileHeldElsewhere(t *testing.T) {
+	db := advisoryTestDB(t)
+	ctx := context.Background()
+
+	// Hold the lock on a connection of our own, standing in for another
+	// replica mid-pass.
+	holder, err := db.Conn(ctx)
+	if err != nil {
+		t.Fatalf("acquire holder connection: %v", err)
+	}
+	defer holder.Close()
+	if _, err := holder.ExecContext(ctx, "SELECT pg_advisory_lock($1)", archiverLockKey); err != nil {
+		t.Fatalf("pre-acquire lock: %v", err)
+	}
+
+	ran := false
+	withAdvisoryLock(ctx, db, archiverLockKey, "test", func() { ran = true })
+	if ran {
+		t.Fatal("fn ran although another connection already held the lock")
+	}
+
+	if _, err := holder.ExecContext(ctx, "SELECT pg_advisory_unlock($1)", archiverLockKey); err != nil {
+		t.Fatalf("release held lock: %v", err)
+	}
+
+	// Now that the holder released it, the next caller should get it.
+	ran = false
+	withAdvisoryLock(ctx, db, archiverLockKey, "test", func() { ran = true })
+	if !ran {
+		t.Fatal("fn did not run after the other connection released the lock")
+	}
+}
+
+func TestWithAdvisoryLock_ReleasesAfterFnReturns(t *testing.T) {
+	db := advisoryTestDB(t)
+	ctx := context.Background()
+
+	withAdvisoryLock(ctx, db, notifierLockKey, "test", func() {})
+
+	// If the first call had leaked the lock, this one would see it held and
+	// skip, leaving ran false.
+	ran := false
+	withAdvisoryLock(ctx, db, notifierLockKey, "test", func() { ran = true })
+	if !ran {
+		t.Fatal("fn did not run on a later call: the earlier call leaked its lock")
+	}
+}
+
+func TestWithAdvisoryLock_KeysAreIndependent(t *testing.T) {
+	db := advisoryTestDB(t)
+	ctx := context.Background()
+
+	holder, err := db.Conn(ctx)
+	if err != nil {
+		t.Fatalf("acquire holder connection: %v", err)
+	}
+	defer holder.Close()
+	if _, err := holder.ExecContext(ctx, "SELECT pg_advisory_lock($1)", archiverLockKey); err != nil {
+		t.Fatalf("pre-acquire archiver lock: %v", err)
+	}
+	defer holder.ExecContext(ctx, "SELECT pg_advisory_unlock($1)", archiverLockKey)
+
+	// Holding archiverLockKey must not block notifierLockKey.
+	ran := false
+	withAdvisoryLock(ctx, db, notifierLockKey, "test", func() { ran = true })
+	if !ran {
+		t.Fatal("fn did not run under a different key although only archiverLockKey was held")
+	}
+}

internal/api/alertmanager.go

@@ -94,10 +94,16 @@ func handleIntegrationWebhook(db *sql.DB, notify NotifyConfig) http.HandlerFunc
 	}
 }
 
+// maxWebhookBodyBytes is larger than maxBodyBytes: a real Alertmanager batch
+// can carry many alerts, each with several labels and annotations, and the
+// sender is a trusted piece of infrastructure rather than an arbitrary
+// caller.
+const maxWebhookBodyBytes = 8 << 20
+
 func receiveWebhook(w http.ResponseWriter, r *http.Request, db *sql.DB, notify NotifyConfig, src alertSource) {
 	teamID := src.teamID
 	var payload amPayload
-	if err := decodeJSON(r, &payload); err != nil {
+	if err := decodeJSONLimit(r, &payload, maxWebhookBodyBytes); err != nil {
 		respond(w, http.StatusBadRequest, errResp("invalid payload"))
 		return
 	}
@@ -169,7 +175,7 @@ func ingest(ctx context.Context, db *sql.DB, notify NotifyConfig, src alertSourc
 		}
 		touched[id] = true
 		alertID := a.id
-		if err := logEvent(ctx, tx, id, evAlertResolved, nil, &alertID, nil); err != nil {
+		if err := logEvent(ctx, tx, id, evAlertResolved, nil, nil, &alertID, nil); err != nil {
 			return err
 		}
 	}
@@ -376,6 +382,17 @@ func incidentForGroup(ctx context.Context, tx *sql.Tx, notify NotifyConfig, team
 // its own. Hence the querier rather than a *sql.Tx. A nil severity leaves the
 // column for refreshSeverity to fill from the member alerts; the sweeper passes
 // one because its incidents have no members to derive it from.
+//
+// Both callers get here only after their own SELECT found no open incident for
+// this group_key — but on more than one replica, two webhook deliveries for the
+// very first occurrence of a brand-new group_key can both pass that SELECT
+// before either INSERTs. ON CONFLICT DO NOTHING against
+// incidents_open_group_key_idx is what makes the loser's INSERT a no-op instead
+// of a unique-violation error that would otherwise roll back its entire
+// payload; existingOpenIncident then hands it the winner's row. Postgres
+// resolves that conflict only once the winner's transaction has committed (or
+// rolled back), so by the time this RETURNING comes back empty, the winner's
+// row is guaranteed visible to that follow-up SELECT.
 func openIncident(ctx context.Context, q querier, notify NotifyConfig, teamID int64, groupKey, title string, groupLabels map[string]string, severity *string) (int64, error) {
 	onCall, err := currentOnCall(ctx, q, teamID)
 	if err != nil {
@@ -391,19 +408,27 @@ func openIncident(ctx context.Context, q querier, notify NotifyConfig, teamID in
 	err = q.QueryRowContext(ctx, `
 		INSERT INTO incidents (team_id, group_key, title, group_labels, signature, status, severity, triggered_at, assigned_to)
 		VALUES ($1, $2, $3, $4::jsonb, $5, 'triggered', $6, $7, $8)
+		ON CONFLICT (team_id, group_key) WHERE resolved_at IS NULL DO NOTHING
 		RETURNING id`,
 		teamID, groupKey, title, string(labelsJSON), incidentSignature(groupLabels, title), severity,
 		time.Now().Unix(), onCall).Scan(&id)
-	if err != nil {
+	switch {
+	case err == sql.ErrNoRows:
+		// Lost the race: someone else's incident for this group_key exists now.
+		// Everything below — the trigger event, assignment, page, escalation
+		// clock — already happened for that row when it was created; attach to
+		// it rather than fail this call (and the whole payload) outright.
+		return existingOpenIncident(ctx, q, teamID, groupKey)
+	case err != nil:
 		return 0, err
 	}
 
-	if err := logEvent(ctx, q, id, evTriggered, nil, nil, nil); err != nil {
+	if err := logEvent(ctx, q, id, evTriggered, nil, nil, nil, nil); err != nil {
 		return 0, err
 	}
 	if onCall != nil {
 		// On an "assigned" event user_id is the assignee, not the actor.
-		if err := logEvent(ctx, q, id, evAssigned, onCall, nil, nil); err != nil {
+		if err := logEvent(ctx, q, id, evAssigned, onCall, nil, nil, nil); err != nil {
 			return 0, err
 		}
 	}
@@ -423,6 +448,21 @@ func openIncident(ctx context.Context, q querier, notify NotifyConfig, teamID in
 	return id, nil
 }
 
+// existingOpenIncident looks up the open incident openIncident's own INSERT just
+// lost a conflict against — the same lookup incidentForGroup does before ever
+// calling openIncident, repeated here for the caller that arrived second.
+func existingOpenIncident(ctx context.Context, q querier, teamID int64, groupKey string) (int64, error) {
+	var id int64
+	err := q.QueryRowContext(ctx,
+		"SELECT id FROM incidents WHERE team_id = $1 AND group_key = $2 AND resolved_at IS NULL",
+		teamID, groupKey,
+	).Scan(&id)
+	if err != nil {
+		return 0, err
+	}
+	return id, nil
+}
+
 // linkAlert adds an alert to an incident, emitting a timeline entry only the
 // first time. Re-sends of an already-linked alert are silent.
 func linkAlert(ctx context.Context, tx *sql.Tx, incidentID, alertID int64) error {
@@ -436,5 +476,5 @@ func linkAlert(ctx context.Context, tx *sql.Tx, incidentID, alertID int64) error
 	if n, _ := res.RowsAffected(); n == 0 {
 		return nil
 	}
-	return logEvent(ctx, tx, incidentID, evAlertAdded, nil, &alertID, nil)
+	return logEvent(ctx, tx, incidentID, evAlertAdded, nil, nil, &alertID, nil)
 }

internal/api/alertmanager_test.go

@@ -0,0 +1,112 @@
+package api_test
+
+import (
+	"bytes"
+	"encoding/json"
+	"fmt"
+	"net/http"
+	"sync"
+	"testing"
+)
+
+// TestWebhook_ConcurrentFirstOccurrenceOpensOneIncident reproduces two
+// replicas racing the very first webhook delivery for a brand-new group_key:
+// both see no open incident yet (incidentForGroup's own SELECT finds
+// nothing) and race openIncident's INSERT.
+//
+// The DB's own unique index already guarantees at most one incident either
+// way, with or without this fix — so "exactly one incident" alone cannot
+// tell a fixed run from a broken one. What ON CONFLICT handling actually
+// changes is what happens to the *loser*: before it, the loser's INSERT hit
+// incidents_open_group_key_idx's unique violation, which — since
+// upsertAlerts ran earlier in that same transaction — rolled back its whole
+// payload, alert insert included. ingest's error is only logged and
+// receiveWebhook answers 200 regardless, so nothing ever retried it: the
+// loser's alert silently never existed. That is the regression signal this
+// test checks — every caller's fingerprint must show up in /api/alerts, not
+// just the winner's.
+func TestWebhook_ConcurrentFirstOccurrenceOpensOneIncident(t *testing.T) {
+	s := newTS(t)
+
+	const callers = 8
+	const groupKey = "race-group"
+
+	// Every caller needs its own fingerprint. A shared one would serialize all
+	// of them at upsertAlerts' own ON CONFLICT (team_id, fingerprint) row lock,
+	// long before any of them reached incidentForGroup — which would hide the
+	// very race this test exists to force.
+	bodies := make([][]byte, callers)
+	for i := range callers {
+		payload := map[string]any{
+			"version":     "4",
+			"status":      "firing",
+			"groupKey":    groupKey,
+			"groupLabels": map[string]string{"alertname": "RaceAlert"},
+			"alerts": []map[string]any{amAlert(fmt.Sprintf("fp-race-%d", i), "RaceAlert", "firing",
+				"2026-05-20T10:00:00Z", "0001-01-01T00:00:00Z", nil)},
+		}
+		bodies[i], _ = json.Marshal(payload)
+	}
+
+	// A start line, so every request is fired as close to simultaneously as
+	// goroutine scheduling allows, rather than trickling out one dial at a
+	// time — the race window is the gap between incidentForGroup's SELECT and
+	// openIncident's INSERT, which a staggered start could easily miss.
+	var ready sync.WaitGroup
+	start := make(chan struct{})
+	statuses := make([]int, callers)
+	var wg sync.WaitGroup
+	for i := range callers {
+		ready.Add(1)
+		wg.Add(1)
+		go func(i int) {
+			defer wg.Done()
+			ready.Done()
+			<-start
+			resp, err := http.Post(s.URL+"/api/integrations/"+s.ingestKey+"/alertmanager",
+				"application/json", bytes.NewReader(bodies[i]))
+			if err != nil {
+				t.Errorf("post webhook #%d: %v", i, err)
+				return
+			}
+			defer resp.Body.Close()
+			statuses[i] = resp.StatusCode
+		}(i)
+	}
+	ready.Wait()
+	close(start)
+	wg.Wait()
+
+	for i, code := range statuses {
+		if code != http.StatusOK {
+			t.Errorf("webhook #%d returned %d, want 200", i, code)
+		}
+	}
+
+	var matched []any
+	for _, inc := range listIncidents(t, s, "") {
+		if inc["group_key"] == groupKey {
+			matched = append(matched, inc["id"])
+		}
+	}
+	if len(matched) != 1 {
+		t.Fatalf("expected exactly 1 incident for group_key %q after %d concurrent deliveries, got %d: %v",
+			groupKey, callers, len(matched), matched)
+	}
+
+	var alerts []map[string]any
+	decode(t, s.req(t, http.MethodGet, "/api/alerts", nil), &alerts)
+	seen := map[string]bool{}
+	for _, a := range alerts {
+		if fp, ok := a["fingerprint"].(string); ok {
+			seen[fp] = true
+		}
+	}
+	for i := range callers {
+		fp := fmt.Sprintf("fp-race-%d", i)
+		if !seen[fp] {
+			t.Errorf("alert %q is missing: its delivery's whole payload was silently rolled back "+
+				"when it lost the race for the incident", fp)
+		}
+	}
+}

internal/api/api_keys_test.go

@@ -0,0 +1,135 @@
+package api_test
+
+import (
+	"net/http"
+	"testing"
+	"time"
+)
+
+func TestAPIKey_DefaultsToNeverExpiring(t *testing.T) {
+	s := newTS(t)
+
+	var key struct {
+		Key       string  `json:"key"`
+		ExpiresAt *string `json:"expires_at"`
+	}
+	decode(t, s.req(t, http.MethodPost, "/api/users/1/api-keys",
+		map[string]string{"name": "no-expiry"}), &key)
+
+	if key.ExpiresAt != nil {
+		t.Errorf("expires_at = %v, want nil (unset expires_in_days means never expires)", *key.ExpiresAt)
+	}
+}
+
+func TestAPIKey_ExpiresInDaysSetsExpiresAt(t *testing.T) {
+	s := newTS(t)
+
+	var key struct {
+		ID        int64   `json:"id"`
+		ExpiresAt *string `json:"expires_at"`
+	}
+	decode(t, s.req(t, http.MethodPost, "/api/users/1/api-keys",
+		map[string]any{"name": "rotates", "expires_in_days": 30}), &key)
+
+	if key.ExpiresAt == nil {
+		t.Fatal("expires_at = nil, want a timestamp roughly 30 days out")
+	}
+	got, err := time.Parse(time.RFC3339, *key.ExpiresAt)
+	if err != nil {
+		t.Fatalf("parse expires_at: %v", err)
+	}
+	want := time.Now().AddDate(0, 0, 30)
+	if diff := want.Sub(got).Abs(); diff > time.Hour {
+		t.Errorf("expires_at = %v, want close to %v (30 days out)", got, want)
+	}
+}
+
+func TestAPIKey_ExpiresInDaysRejectsOutOfRange(t *testing.T) {
+	s := newTS(t)
+
+	for _, days := range []int{-1, 3651} {
+		resp := s.req(t, http.MethodPost, "/api/users/1/api-keys",
+			map[string]any{"name": "bad", "expires_in_days": days})
+		resp.Body.Close()
+		if resp.StatusCode != http.StatusBadRequest {
+			t.Errorf("expires_in_days=%d: status = %d, want %d", days, resp.StatusCode, http.StatusBadRequest)
+		}
+	}
+}
+
+func TestAPIKey_AnExpiredKeyCannotAuthenticate(t *testing.T) {
+	s := newTS(t)
+
+	var key struct {
+		ID  int64  `json:"id"`
+		Key string `json:"key"`
+	}
+	decode(t, s.req(t, http.MethodPost, "/api/users/1/api-keys",
+		map[string]any{"name": "soon-expired", "expires_in_days": 1}), &key)
+
+	// A fresh key works...
+	req, _ := http.NewRequest(http.MethodGet, s.URL+"/api/me", nil)
+	req.Header.Set("Authorization", "Bearer "+key.Key)
+	resp, err := http.DefaultClient.Do(req)
+	if err != nil {
+		t.Fatalf("GET /api/me: %v", err)
+	}
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("fresh key: status = %d, want %d", resp.StatusCode, http.StatusOK)
+	}
+
+	// ...and stops working once its expiry has passed.
+	s.exec(t, "UPDATE api_keys SET expires_at = $1 WHERE id = $2", time.Now().Add(-time.Hour).Unix(), key.ID)
+
+	req2, _ := http.NewRequest(http.MethodGet, s.URL+"/api/me", nil)
+	req2.Header.Set("Authorization", "Bearer "+key.Key)
+	resp2, err := http.DefaultClient.Do(req2)
+	if err != nil {
+		t.Fatalf("GET /api/me: %v", err)
+	}
+	defer resp2.Body.Close()
+	if resp2.StatusCode != http.StatusUnauthorized {
+		t.Errorf("expired key: status = %d, want %d", resp2.StatusCode, http.StatusUnauthorized)
+	}
+}
+
+func TestAPIKey_ListNeverReturnsTheRawKey(t *testing.T) {
+	s := newTS(t)
+
+	decode(t, s.req(t, http.MethodPost, "/api/users/1/api-keys",
+		map[string]string{"name": "listed"}), new(struct {
+		Key string `json:"key"`
+	}))
+
+	var keys []struct {
+		ID   int64  `json:"id"`
+		Name string `json:"name"`
+		Key  string `json:"key"`
+	}
+	decode(t, s.req(t, http.MethodGet, "/api/users/1/api-keys", nil), &keys)
+
+	found := false
+	for _, k := range keys {
+		if k.Name == "listed" {
+			found = true
+		}
+		if k.Key != "" {
+			t.Errorf("key %d (%s): raw key present in listing", k.ID, k.Name)
+		}
+	}
+	if !found {
+		t.Error("the key just created does not appear in the listing")
+	}
+}
+
+func TestAPIKey_ListIsSelfOrAdmin(t *testing.T) {
+	s := newTS(t)
+	a := newTeam(t, s, "apikeys-a")
+
+	resp := a.call(http.MethodGet, "/api/users/1/api-keys", nil)
+	defer resp.Body.Close()
+	if resp.StatusCode != http.StatusForbidden {
+		t.Errorf("status = %d, want %d (not self, not an admin)", resp.StatusCode, http.StatusForbidden)
+	}
+}

internal/api/archiver.go

@@ -15,6 +15,12 @@ const (
 	// expiryGrace absorbs clock skew and notification latency before an alert
 	// whose ends_at watermark has passed is treated as stale.
 	expiryGrace = 5 * time.Minute
+
+	// archiverLockKey is the Postgres advisory lock the sweeper takes for the
+	// duration of each pass, so that running more than one replica does not run
+	// the sweep concurrently on all of them. Its value has no meaning beyond
+	// being distinct from notifierLockKey.
+	archiverLockKey int64 = 7265_0001
 )
 
 // StartArchiver runs the alert sweeper until ctx is cancelled, starting with an
@@ -23,15 +29,25 @@ const (
 // the fallback, not the setting: each pass reads the current value from the
 // settings table, so an administrator's change takes effect on the next tick
 // instead of at the next restart.
+//
+// Each pass runs under archiverLockKey (see withAdvisoryLock), so that on more
+// than one replica only whichever instance's tick takes the lock first actually
+// sweeps; the rest skip that tick rather than racing the same pass.
 func StartArchiver(ctx context.Context, db *sql.DB, archiveAfter, staleAfter time.Duration, notify NotifyConfig) {
 	ticker := time.NewTicker(sweepInterval)
 	defer ticker.Stop()
 
-	Sweep(ctx, db, archiveAfter, staleAfter, notify)
+	sweep := func() {
+		withAdvisoryLock(ctx, db, archiverLockKey, "sweeper", func() {
+			Sweep(ctx, db, archiveAfter, staleAfter, notify)
+		})
+	}
+
+	sweep()
 	for {
 		select {
 		case <-ticker.C:
-			Sweep(ctx, db, archiveAfter, staleAfter, notify)
+			sweep()
 		case <-ctx.Done():
 			return
 		}
@@ -60,6 +76,7 @@ func Sweep(ctx context.Context, db *sql.DB, archiveAfter, staleAfter time.Durati
 	archiveResolvedIncidents(ctx, db, archiveAfter)
 	purgeAckTokens(ctx, db)
 	purgeSessions(ctx, db)
+	purgeRateLimits(ctx, db)
 }
 
 // expireStale resolves firing alerts that Alertmanager has stopped refreshing.
@@ -105,6 +122,10 @@ func expireStale(ctx context.Context, db *sql.DB, staleAfter time.Duration, skip
 	for i, id := range ids {
 		idList[i] = id
 	}
+	// #nosec G202 -- sqlArgs.add/addList only ever splice in the "$N"
+	// placeholder they hand back, never a value; every value travels through
+	// args.all() as a bound parameter. See the sqlArgs doc comment in
+	// helpers.go.
 	if _, err := db.ExecContext(ctx, `
 		UPDATE alerts
 		   SET status            = 'resolved',
@@ -126,7 +147,7 @@ func expireStale(ctx context.Context, db *sql.DB, staleAfter time.Duration, skip
 			continue
 		}
 		alertID := id
-		if err := logEvent(ctx, db, incidentID, evAlertResolved, nil, &alertID, nil); err != nil {
+		if err := logEvent(ctx, db, incidentID, evAlertResolved, nil, nil, &alertID, nil); err != nil {
 			log.Printf("sweeper: log expiry event: %v", err)
 		}
 	}

internal/api/auth.go

@@ -45,57 +45,85 @@ var dummyHash = sync.OnceValue(func() []byte {
 	return h
 })
 
-// loginLimiter counts failed logins in a fixed window, per username and per
-// client address. The username limit is what stops guessing one account; the
-// address limit is looser because every user behind the same gateway or NAT
-// shares it.
+// loginLimiter counts failed logins (and other unauthenticated attempts:
+// sign-up, OIDC/device start) in a fixed window, per key — a username, a
+// client address, or both, depending on the caller.
+//
+// Backed by Postgres rather than an in-memory map: this server runs more
+// than one replica in production (v0.37.0), and a counter that only ever
+// sees its own pod's traffic would quietly let every limit through
+// multiplied by the replica count — two loginLimiter values pointed at the
+// same db, standing in for two replicas, now share exactly one count per
+// key instead of each keeping their own.
+//
+// The window resets rather than slides, the same behavior the in-memory
+// version it replaces had: once a key's window is older than loginWindow,
+// the next fail() starts a fresh one instead of extending the stale one.
 type loginLimiter struct {
-	mu       sync.Mutex
-	failures map[string]*loginWindowCount
+	db *sql.DB
 }
 
-type loginWindowCount struct {
-	start time.Time
-	n     int
+func newLoginLimiter(db *sql.DB) *loginLimiter {
+	return &loginLimiter{db: db}
 }
 
-func newLoginLimiter() *loginLimiter {
-	return &loginLimiter{failures: map[string]*loginWindowCount{}}
-}
-
-func (l *loginLimiter) blocked(key string, max int) bool {
-	l.mu.Lock()
-	defer l.mu.Unlock()
-	c, ok := l.failures[key]
-	if !ok || time.Since(c.start) > loginWindow {
+func (l *loginLimiter) blocked(ctx context.Context, key string, max int) bool {
+	cutoff := time.Now().Unix() - int64(loginWindow.Seconds())
+	var count int
+	err := l.db.QueryRowContext(ctx, `
+		SELECT count FROM rate_limit_counters
+		 WHERE key = $1 AND window_start > $2`,
+		key, cutoff,
+	).Scan(&count)
+	if err != nil {
+		// No row (never failed, or its window already expired): not blocked.
+		// A real query error fails the same way — a rate limiter that locks
+		// everyone out during a brief database hiccup is worse than one that
+		// is briefly too generous.
 		return false
 	}
-	return c.n >= max
+	return count >= max
 }
 
-func (l *loginLimiter) fail(keys ...string) {
-	l.mu.Lock()
-	defer l.mu.Unlock()
-	now := time.Now()
-	for k, c := range l.failures {
-		if now.Sub(c.start) > loginWindow {
-			delete(l.failures, k)
-		}
-	}
+func (l *loginLimiter) fail(ctx context.Context, keys ...string) {
+	now := time.Now().Unix()
+	windowSecs := int64(loginWindow.Seconds())
 	for _, key := range keys {
-		c, ok := l.failures[key]
-		if !ok {
-			c = &loginWindowCount{start: now}
-			l.failures[key] = c
+		if _, err := l.db.ExecContext(ctx, `
+			INSERT INTO rate_limit_counters (key, window_start, count)
+			VALUES ($1, $2, 1)
+			ON CONFLICT (key) DO UPDATE SET
+				window_start = CASE WHEN rate_limit_counters.window_start <= $2 - $3
+				                    THEN $2 ELSE rate_limit_counters.window_start END,
+				count        = CASE WHEN rate_limit_counters.window_start <= $2 - $3
+				                    THEN 1 ELSE rate_limit_counters.count + 1 END`,
+			key, now, windowSecs,
+		); err != nil {
+			log.Printf("rate limiter: record failure for %q: %v", key, err) // #nosec G706 -- %q
 		}
-		c.n++
 	}
 }
 
-func (l *loginLimiter) clear(key string) {
-	l.mu.Lock()
-	defer l.mu.Unlock()
-	delete(l.failures, key)
+func (l *loginLimiter) clear(ctx context.Context, key string) {
+	if _, err := l.db.ExecContext(ctx, "DELETE FROM rate_limit_counters WHERE key = $1", key); err != nil {
+		log.Printf("rate limiter: clear %q: %v", key, err)
+	}
+}
+
+// purgeRateLimits deletes rate-limit windows that have expired, from the
+// sweeper — otherwise every distinct username and address this server has
+// ever seen a failed attempt from would stay a row forever.
+func purgeRateLimits(ctx context.Context, db *sql.DB) {
+	cutoff := time.Now().Unix() - int64(loginWindow.Seconds())
+	res, err := db.ExecContext(ctx,
+		"DELETE FROM rate_limit_counters WHERE window_start <= $1", cutoff)
+	if err != nil {
+		log.Printf("sweeper: purge rate limit counters: %v", err)
+		return
+	}
+	if n, _ := res.RowsAffected(); n > 0 {
+		log.Printf("sweeper: purged %d expired rate limit counter(s)", n)
+	}
 }
 
 // clientAddr is the address a login is counted against. Behind the gateway
@@ -171,6 +199,9 @@ func startSessionCapped(w http.ResponseWriter, r *http.Request, db *sql.DB, user
 		return err
 	}
 
+	// #nosec G124 -- HttpOnly/SameSite are literal below; Secure is
+	// cookieSecure(publicURL, r), not a literal true, which is what trips
+	// this rule. See cookieSecure's own doc comment above.
 	http.SetCookie(w, &http.Cookie{
 		Name:     sessionCookie,
 		Value:    raw,
@@ -198,7 +229,7 @@ func handleLogin(db *sql.DB, limiter *loginLimiter, publicURL string) http.Handl
 		userKey := "user:" + strings.ToLower(username)
 		addrKey := "addr:" + clientAddr(r)
 
-		if limiter.blocked(userKey, loginMaxPerUser) || limiter.blocked(addrKey, loginMaxPerAddr) {
+		if limiter.blocked(r.Context(), userKey, loginMaxPerUser) || limiter.blocked(r.Context(), addrKey, loginMaxPerAddr) {
 			w.Header().Set("Retry-After", strconv.Itoa(int(loginWindow.Seconds())))
 			respond(w, http.StatusTooManyRequests, errResp("too many failed attempts, try again later"))
 			return
@@ -220,11 +251,11 @@ func handleLogin(db *sql.DB, limiter *loginLimiter, publicURL string) http.Handl
 		}
 		match := bcrypt.CompareHashAndPassword(stored, []byte(req.Password)) == nil
 		if !match || !hash.Valid {
-			limiter.fail(userKey, addrKey)
+			limiter.fail(r.Context(), userKey, addrKey)
 			respond(w, http.StatusUnauthorized, errResp("invalid username or password"))
 			return
 		}
-		limiter.clear(userKey)
+		limiter.clear(r.Context(), userKey)
 
 		if err := startSession(w, r, db, userID, publicURL); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
@@ -252,6 +283,9 @@ func handleLogout(db *sql.DB, publicURL string) http.HandlerFunc {
 		if c, err := r.Cookie(sessionCookie); err == nil && c.Value != "" {
 			db.ExecContext(r.Context(), "DELETE FROM sessions WHERE token_hash = $1", hashToken(c.Value))
 		}
+		// #nosec G124 -- HttpOnly/SameSite are literal below; Secure is
+		// cookieSecure(publicURL, r), not a literal true, which is what
+		// trips this rule. See cookieSecure's own doc comment above.
 		http.SetCookie(w, &http.Cookie{
 			Name:     sessionCookie,
 			Value:    "",
@@ -291,9 +325,16 @@ func handleMe(db *sql.DB) http.HandlerFunc {
 		}
 		var hash sql.NullString
 		var dismissed *int64
-		db.QueryRowContext(r.Context(),
+		if err := db.QueryRowContext(r.Context(),
 			"SELECT password_hash, onboarding_dismissed_at FROM users WHERE id = $1",
-			caller.ID).Scan(&hash, &dismissed)
+			caller.ID).Scan(&hash, &dismissed); err != nil {
+			// fetchUser above already found this row, so an error here is a
+			// transient database problem, not a missing user — worth a 500
+			// rather than silently answering "no password, not dismissed",
+			// which a client would otherwise take at face value.
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		respond(w, http.StatusOK, meResponse{
 			User:                user,
 			HasPassword:         hash.Valid,

internal/api/authz_scope_test.go

@@ -0,0 +1,182 @@
+package api_test
+
+import (
+	"net/http"
+	"testing"
+)
+
+// This file is the regression test for the pattern documented throughout
+// middleware.go: every team-scoped handler calls requireTeamMember or
+// requireTeamOwner before touching data, every self-or-admin handler calls
+// requireSelfOrAdmin, and every admin-only route sits behind AdminOnly. That
+// pattern is enforced by convention, not by the type system — a new handler
+// that forgets the call would compile and pass review on a quick read just
+// as easily as one that remembers it. These tests exercise every route that
+// carries one of those guards as a caller who should be refused, so a future
+// handler missing its guard fails CI instead of becoming a silent IDOR.
+
+// TestAuthzScope_TeamScopedRoutesRefuseANonMember builds two teams and, for
+// every team-scoped route, calls it as team A's owner against team B's
+// resources. requireTeamMember and requireTeamOwner both answer a non-member
+// with 404 (team.go's own reasoning: whether a team exists is itself
+// something only its members should learn), so every one of these must come
+// back 404 regardless of which of the two guards its handler uses.
+func TestAuthzScope_TeamScopedRoutesRefuseANonMember(t *testing.T) {
+	s := newTS(t)
+	a := newTeam(t, s, "authz-a")
+	b := newTeam(t, s, "authz-b")
+
+	// An incident in B, to cover the ID-based routes under /api/incidents —
+	// scoped by the incident's own team_id rather than a {teamID} path
+	// segment, but through the same single chokepoint (incidentIDParam).
+	postToIntegration(t, s, b.key, "fp-authz-scope", "AuthzScopeAlert")
+	var incidents []struct {
+		ID int64 `json:"id"`
+	}
+	decode(t, b.call(http.MethodGet, "/api/incidents", nil), &incidents)
+	if len(incidents) == 0 {
+		t.Fatal("setup: no incident in team B to test against")
+	}
+	incidentPath := "/api/incidents/" + id64(incidents[0].ID)
+
+	bPath := "/api/teams/" + id64(b.id)
+	tests := []struct {
+		method, path string
+	}{
+		// Team membership/ownership itself.
+		{http.MethodPut, bPath},
+		{http.MethodDelete, bPath},
+		{http.MethodGet, bPath + "/members"},
+		{http.MethodPost, bPath + "/members"},
+		{http.MethodDelete, bPath + "/members/1"},
+
+		// OIDC group binding.
+		{http.MethodGet, bPath + "/oidc-groups"},
+		{http.MethodPut, bPath + "/oidc-groups"},
+
+		// Invites.
+		{http.MethodGet, bPath + "/invites"},
+		{http.MethodPost, bPath + "/invites"},
+		{http.MethodDelete, bPath + "/invites/1"},
+
+		// Escalation.
+		{http.MethodGet, bPath + "/escalation"},
+		{http.MethodPut, bPath + "/escalation"},
+
+		// Dead man's switches.
+		{http.MethodGet, bPath + "/deadman/switches"},
+		{http.MethodPost, bPath + "/deadman/switches"},
+		{http.MethodPut, bPath + "/deadman/switches/1"},
+		{http.MethodDelete, bPath + "/deadman/switches/1"},
+
+		// Integrations.
+		{http.MethodGet, bPath + "/integrations"},
+		{http.MethodPost, bPath + "/integrations"},
+		{http.MethodPatch, bPath + "/integrations/1"},
+		{http.MethodDelete, bPath + "/integrations/1"},
+
+		// Schedule.
+		{http.MethodGet, bPath + "/schedule"},
+		{http.MethodPost, bPath + "/schedule"},
+		{http.MethodDelete, bPath + "/schedule/1"},
+
+		// Incidents, scoped by the incident's own team rather than a
+		// {teamID} segment.
+		{http.MethodGet, incidentPath},
+		{http.MethodGet, incidentPath + "/alerts"},
+		{http.MethodGet, incidentPath + "/timeline"},
+		{http.MethodGet, incidentPath + "/similar"},
+		{http.MethodPost, incidentPath + "/acknowledge"},
+		{http.MethodDelete, incidentPath + "/acknowledge"},
+		{http.MethodPost, incidentPath + "/resolve"},
+		{http.MethodPost, incidentPath + "/assign"},
+		{http.MethodPost, incidentPath + "/snooze"},
+		{http.MethodDelete, incidentPath + "/snooze"},
+		{http.MethodPost, incidentPath + "/archive"},
+		{http.MethodDelete, incidentPath + "/archive"},
+		{http.MethodPost, incidentPath + "/notes"},
+		{http.MethodDelete, incidentPath + "/notes/1"},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.method+" "+tc.path, func(t *testing.T) {
+			resp := a.call(tc.method, tc.path, nil)
+			defer resp.Body.Close()
+			if resp.StatusCode != http.StatusNotFound {
+				t.Errorf("status = %d, want %d (A is not a member of B)", resp.StatusCode, http.StatusNotFound)
+			}
+		})
+	}
+}
+
+// TestAuthzScope_AdminOnlyRoutesRefuseANonAdmin exercises AdminOnly's group
+// in router.go directly: a signed-in, non-admin caller gets 403 from every
+// route in it, before any handler body runs.
+func TestAuthzScope_AdminOnlyRoutesRefuseANonAdmin(t *testing.T) {
+	s := newTS(t)
+	a := newTeam(t, s, "authz-admin")
+
+	tests := []struct {
+		method, path string
+	}{
+		{http.MethodPost, "/api/users"},
+		{http.MethodDelete, "/api/users/1"},
+		{http.MethodPut, "/api/users/1/admin"},
+		{http.MethodPut, "/api/users/1/disabled"},
+		{http.MethodGet, "/api/admin/teams"},
+		{http.MethodGet, "/api/admin/teams/" + id64(a.id)},
+		{http.MethodGet, "/api/admin/settings"},
+		{http.MethodPut, "/api/admin/settings"},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.method+" "+tc.path, func(t *testing.T) {
+			resp := a.call(tc.method, tc.path, nil)
+			defer resp.Body.Close()
+			if resp.StatusCode != http.StatusForbidden {
+				t.Errorf("status = %d, want %d (not an admin)", resp.StatusCode, http.StatusForbidden)
+			}
+		})
+	}
+}
+
+// TestAuthzScope_SelfOrAdminRoutesRefuseAnotherNonAdminUser exercises
+// requireSelfOrAdmin's call sites: a non-admin caller acting on a *different*
+// user's account must be refused, the same as AdminOnly's routes, even
+// though these sit in the general authenticated group rather than behind
+// AdminOnly itself.
+func TestAuthzScope_SelfOrAdminRoutesRefuseAnotherNonAdminUser(t *testing.T) {
+	s := newTS(t)
+	a := newTeam(t, s, "authz-self-a")
+	b := newTeam(t, s, "authz-self-b")
+
+	var members []struct {
+		UserID int64 `json:"user_id"`
+	}
+	decode(t, s.req(t, http.MethodGet, "/api/teams/"+id64(b.id)+"/members", nil), &members)
+	if len(members) == 0 {
+		t.Fatal("setup: team B has no members")
+	}
+	bUserID := id64(members[0].UserID)
+
+	tests := []struct {
+		method, path string
+	}{
+		{http.MethodGet, "/api/users/" + bUserID + "/teams"},
+		{http.MethodPut, "/api/users/" + bUserID + "/notify"},
+		{http.MethodPut, "/api/users/" + bUserID + "/password"},
+		{http.MethodGet, "/api/users/" + bUserID + "/api-keys"},
+		{http.MethodPost, "/api/users/" + bUserID + "/api-keys"},
+		{http.MethodDelete, "/api/users/" + bUserID + "/api-keys/1"},
+	}
+
+	for _, tc := range tests {
+		t.Run(tc.method+" "+tc.path, func(t *testing.T) {
+			resp := a.call(tc.method, tc.path, nil)
+			defer resp.Body.Close()
+			if resp.StatusCode != http.StatusForbidden {
+				t.Errorf("status = %d, want %d (not self, not an admin)", resp.StatusCode, http.StatusForbidden)
+			}
+		})
+	}
+}

internal/api/caller.go

@@ -99,13 +99,24 @@ func (c Caller) ServiceAccountID() (int64, bool) {
 	return c.sa.id, true
 }
 
+// ServiceAccountName reports this caller's own service-account name, for a
+// handler's synchronous response — the same credential it authenticated
+// with, already resolved onto the Caller by serveAsServiceAccount, so no
+// extra query is needed.
+func (c Caller) ServiceAccountName() (string, bool) {
+	if c.sa == nil {
+		return "", false
+	}
+	return c.sa.name, true
+}
+
 // Identity is a stable, log/audit-facing string distinguishing a human
 // caller from a service account — "user:42" or "service-account:7". Not
-// wired into any database column today (incidents.go's acknowledged_by/
-// assigned_to/user_id are explicitly out of scope for this change — that
-// needs its own schema migration, tracked separately), but this is the one
-// place in the request path that already knows which kind of caller this
-// is, and that follow-up will want exactly this accessor.
+// wired into any database column — incidents.go's acknowledged_by/
+// incident_events.user_id use AsHuman()/ServiceAccountID() directly against
+// the parallel *_service_account_id columns (migration 015) instead, since a
+// column needs the id, not this rendered string. assigned_to stays
+// human-only and out of scope (terdut-server#25's follow-up).
 func (c Caller) Identity() string {
 	switch {
 	case c.user != nil:

internal/api/deadman.go

@@ -283,6 +283,10 @@ func deadmanAlerts(ctx context.Context, db *sql.DB, teamID int64, cfg deadmanSet
 		nameList[i] = n
 	}
 
+	// #nosec G202 -- sqlArgs.add/addList only ever splice in the "$N"
+	// placeholder they hand back, never a value; every value travels through
+	// args.all() as a bound parameter. See the sqlArgs doc comment in
+	// helpers.go.
 	rows, err := db.QueryContext(ctx, `
 		SELECT id, team_id, fingerprint, labels, status, received_at
 		FROM   alerts
@@ -373,7 +377,7 @@ func deadmanDied(ctx context.Context, db *sql.DB, notify NotifyConfig, hb deadma
 
 	alertID := hb.id
 	detail := "last heartbeat " + humanDuration(now.Sub(time.Unix(hb.receivedAt, 0))) + " ago"
-	if err := logEvent(ctx, tx, incidentID, evDeadmanSilent, nil, &alertID, &detail); err != nil {
+	if err := logEvent(ctx, tx, incidentID, evDeadmanSilent, nil, nil, &alertID, &detail); err != nil {
 		return err
 	}
 
@@ -415,7 +419,7 @@ func deadmanRecovered(ctx context.Context, db *sql.DB, hb deadmanAlert) error {
 		time.Now().Unix(), incidentResolutionRecovered, incidentID); err != nil {
 		return err
 	}
-	if err := logEvent(ctx, tx, incidentID, evResolved, nil, nil, nil); err != nil {
+	if err := logEvent(ctx, tx, incidentID, evResolved, nil, nil, nil, nil); err != nil {
 		return err
 	}
 	// The all-clear goes to whoever was paged, which enqueueResolved works out

internal/api/device.go

@@ -72,12 +72,12 @@ func normalizeUserCode(s string) string {
 func handleDeviceStart(db *sql.DB, limiter *loginLimiter, publicURL string) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		addrKey := "device:" + clientAddr(r)
-		if limiter.blocked(addrKey, deviceStartMaxPerAddr) {
+		if limiter.blocked(r.Context(), addrKey, deviceStartMaxPerAddr) {
 			w.Header().Set("Retry-After", strconv.Itoa(int(loginWindow.Seconds())))
 			respond(w, http.StatusTooManyRequests, errResp("too many sign-in attempts, try again later"))
 			return
 		}
-		limiter.fail(addrKey)
+		limiter.fail(r.Context(), addrKey)
 
 		deviceCode, deviceHash, err := randomToken()
 		if err != nil {

internal/api/escalation.go

@@ -229,7 +229,7 @@ func advanceEscalation(ctx context.Context, db *sql.DB, cfg NotifyConfig, policy
 		// nobody. That is a policy that looks configured and is not.
 		detail += ": nobody reachable"
 	}
-	if err := logEvent(ctx, tx, incidentID, evEscalated, nil, nil, &detail); err != nil {
+	if err := logEvent(ctx, tx, incidentID, evEscalated, nil, nil, nil, &detail); err != nil {
 		return err
 	}
 	return tx.Commit()
@@ -256,7 +256,7 @@ func escalationExhausted(ctx context.Context, tx *sql.Tx, policy *escalationPoli
 		incidentID); err != nil {
 		return err
 	}
-	return logEvent(ctx, tx, incidentID, evEscalated, nil, nil, &detail)
+	return logEvent(ctx, tx, incidentID, evEscalated, nil, nil, nil, &detail)
 }
 
 // pageLevel notifies every target of one level and reports who was woken.

internal/api/helpers.go

@@ -1,8 +1,11 @@
 package api
 
 import (
+	"context"
+	"database/sql"
 	"encoding/json"
 	"errors"
+	"log"
 	"net/http"
 	"strconv"
 	"strings"
@@ -69,11 +72,64 @@ func respond(w http.ResponseWriter, status int, v any) {
 	json.NewEncoder(w).Encode(v)
 }
 
+// maxBodyBytes caps an ordinary JSON request body. 1 MiB is far more than any
+// endpoint below needs — it exists so an unauthenticated caller (signup,
+// login, bootstrap) can't make the server buffer an arbitrarily large body
+// before the request is even validated.
+const maxBodyBytes = 1 << 20
+
 func decodeJSON(r *http.Request, v any) error {
+	return decodeJSONLimit(r, v, maxBodyBytes)
+}
+
+// decodeJSONLimit is decodeJSON with an explicit cap, for the one endpoint
+// (the Alertmanager webhook, see maxWebhookBodyBytes) whose real payloads can
+// legitimately be larger than maxBodyBytes.
+func decodeJSONLimit(r *http.Request, v any, limit int64) error {
 	defer r.Body.Close()
+	// w is nil: there is no ResponseWriter here to disable keep-alive with,
+	// which net/http documents as fine — the limit is still enforced, the
+	// connection just isn't closed early on a request that blows past it.
+	r.Body = http.MaxBytesReader(nil, r.Body, limit)
 	return json.NewDecoder(r.Body).Decode(v)
 }
 
 func errResp(msg string) map[string]string {
 	return map[string]string{"error": msg}
 }
+
+// withAdvisoryLock runs fn only if it can take the named Postgres advisory lock on a
+// dedicated connection, and skips fn otherwise. This is what keeps the archiver and
+// notifier safe to run on more than one replica: whichever instance's tick gets there
+// first does the work; the rest see the lock held and simply wait for their next tick
+// instead of running the same pass concurrently.
+//
+// pg_try_advisory_lock is session-scoped, so taking and releasing it must happen on the
+// same connection, reserved via db.Conn rather than borrowed from the pool's shared
+// connections fn itself may use — and released (unlocked, then closed) before returning,
+// since a session lock otherwise outlives this call and leaks onto whatever reuses the
+// pooled connection next.
+func withAdvisoryLock(ctx context.Context, db *sql.DB, key int64, name string, fn func()) {
+	conn, err := db.Conn(ctx)
+	if err != nil {
+		log.Printf("%s: advisory lock: acquire connection: %v", name, err)
+		return
+	}
+	defer conn.Close()
+
+	var locked bool
+	if err := conn.QueryRowContext(ctx, "SELECT pg_try_advisory_lock($1)", key).Scan(&locked); err != nil {
+		log.Printf("%s: advisory lock: %v", name, err)
+		return
+	}
+	if !locked {
+		return // another replica is already running this pass
+	}
+	defer func() {
+		if _, err := conn.ExecContext(ctx, "SELECT pg_advisory_unlock($1)", key); err != nil {
+			log.Printf("%s: advisory unlock: %v", name, err)
+		}
+	}()
+
+	fn()
+}

internal/api/incident_store.go

@@ -32,6 +32,8 @@ const (
 	evAcknowledged   = "acknowledged"
 	evUnacknowledged = "unacknowledged"
 	evAssigned       = "assigned"
+	evArchived       = "archived"
+	evUnarchived     = "unarchived"
 	evSnoozed        = "snoozed"
 	evUnsnoozed      = "unsnoozed"
 	evResolved       = "resolved"
@@ -65,11 +67,13 @@ const incidentSelectFrom = `
 	         WHERE el.team_id = i.team_id AND el.position = i.escalation_level),
 	       i.triggered_at,
 	       i.acknowledged_by, i.acknowledged_at, ack.username,
+	       i.acknowledged_by_service_account_id, acksa.name,
 	       i.assigned_to, asg.username, i.snoozed_until,
 	       i.resolved_at, i.resolution_source, i.archived_at
 	FROM   incidents i
 	JOIN   teams t ON t.id = i.team_id
 	LEFT   JOIN users ack ON ack.id = i.acknowledged_by
+	LEFT   JOIN service_accounts acksa ON acksa.id = i.acknowledged_by_service_account_id
 	LEFT   JOIN users asg ON asg.id = i.assigned_to`
 
 func scanIncident(s scanner) (models.Incident, error) {
@@ -83,6 +87,7 @@ func scanIncident(s scanner) (models.Incident, error) {
 		&i.EscalationLevel, &escalationDue,
 		&triggeredAt,
 		&i.AcknowledgedByID, &ackAt, &i.AcknowledgedByUser,
+		&i.AcknowledgedByServiceAccountID, &i.AcknowledgedByServiceAccountName,
 		&i.AssignedToID, &i.AssignedToUser, &snoozedUntil,
 		&resolvedAt, &i.ResolutionSource, &archivedAt,
 	); err != nil {
@@ -112,13 +117,42 @@ func fetchIncident(ctx context.Context, q querier, id int64) (models.Incident, e
 	return scanIncident(q.QueryRowContext(ctx, incidentSelectFrom+" WHERE i.id = $1", id))
 }
 
-// logEvent appends one entry to an incident's timeline. A nil userID means the
-// server acted rather than a person.
-func logEvent(ctx context.Context, q querier, incidentID int64, evType string, userID, alertID *int64, detail *string) error {
+// callerActorIDs resolves the current request's caller into the pair of
+// nilable ids logEvent/acknowledgeIncidentAs expect: exactly one of userID/
+// serviceAccountID is set (never both), replacing the unchecked
+// userFromContext(ctx) zero-value reads that used to write a human-only id
+// of 0 for a service-account caller (terdut-server#25).
+func callerActorIDs(ctx context.Context) (userID, serviceAccountID *int64) {
+	caller, _ := callerFromContext(ctx)
+	if u, ok := caller.AsHuman(); ok {
+		return &u.ID, nil
+	}
+	if id, ok := caller.ServiceAccountID(); ok {
+		return nil, &id
+	}
+	return nil, nil
+}
+
+// logEvent appends one entry to an incident's timeline. userID and
+// serviceAccountID are mutually exclusive and both nilable; both nil means
+// the server acted rather than any caller (see incident_events_actor_xor_chk,
+// migration 015).
+func logEvent(ctx context.Context, q querier, incidentID int64, evType string, userID, serviceAccountID, alertID *int64, detail *string) error {
 	_, err := q.ExecContext(ctx, `
-		INSERT INTO incident_events (incident_id, type, user_id, alert_id, detail, created_at)
+		INSERT INTO incident_events (incident_id, type, user_id, service_account_id, alert_id, detail, created_at)
+		VALUES ($1, $2, $3, $4, $5, $6, $7)`,
+		incidentID, evType, userID, serviceAccountID, alertID, detail, time.Now().Unix())
+	return err
+}
+
+// logAssignedEvent records an assignment: user_id is the assignee, and the
+// caller who performed it goes in the actor_* columns (migration 018), since
+// user_id cannot hold both.
+func logAssignedEvent(ctx context.Context, q querier, incidentID, assigneeID int64, actorUserID, actorServiceAccountID *int64) error {
+	_, err := q.ExecContext(ctx, `
+		INSERT INTO incident_events (incident_id, type, user_id, actor_user_id, actor_service_account_id, created_at)
 		VALUES ($1, $2, $3, $4, $5, $6)`,
-		incidentID, evType, userID, alertID, detail, time.Now().Unix())
+		incidentID, evAssigned, assigneeID, actorUserID, actorServiceAccountID, time.Now().Unix())
 	return err
 }
 
@@ -251,7 +285,7 @@ func resolveIfSettled(ctx context.Context, q querier, incidentID int64) (bool, e
 	if err := stopEscalation(ctx, q, incidentID); err != nil {
 		return false, err
 	}
-	if err := logEvent(ctx, q, incidentID, evResolved, nil, nil, nil); err != nil {
+	if err := logEvent(ctx, q, incidentID, evResolved, nil, nil, nil, nil); err != nil {
 		return false, err
 	}
 	// The all-clear goes only to whoever was paged in the first place, which
@@ -260,19 +294,30 @@ func resolveIfSettled(ctx context.Context, q querier, incidentID int64) (bool, e
 	return true, enqueueResolved(ctx, q, incidentID)
 }
 
-// acknowledgeIncident records that userID has picked an incident up, and reports
-// whether it changed anything — an already-resolved or already-acknowledged
-// incident is left alone, so a second acknowledge (a retried request, or a
-// stale push notification tapped after the web UI already acked it) is a
-// no-op rather than a second "acknowledged" timeline entry. Shared by the
-// authenticated handler and the Acknowledge button in a push notification,
-// so both write the same state and the same timeline entry.
+// acknowledgeIncident records that userID — a human — has picked an incident
+// up, and reports whether it changed anything — an already-resolved or
+// already-acknowledged incident is left alone, so a second acknowledge (a
+// retried request, or a stale push notification tapped after the web UI
+// already acked it) is a no-op rather than a second "acknowledged" timeline
+// entry. Used only by the Acknowledge button in a push notification
+// (notify_ack.go), which always resolves a human from
+// incident_ack_tokens.user_id — there is no service-account equivalent of
+// that flow, so this keeps its human-only signature; the authenticated
+// handler goes through acknowledgeIncidentAs below instead.
 func acknowledgeIncident(ctx context.Context, q querier, incidentID, userID int64) (bool, error) {
+	return acknowledgeIncidentAs(ctx, q, incidentID, &userID, nil)
+}
+
+// acknowledgeIncidentAs is acknowledgeIncident generalized to either actor
+// kind. userID and serviceAccountID are mutually exclusive and nilable the
+// same way logEvent's are (see incidents_ack_actor_xor_chk, migration 015).
+func acknowledgeIncidentAs(ctx context.Context, q querier, incidentID int64, userID, serviceAccountID *int64) (bool, error) {
 	res, err := q.ExecContext(ctx, `
 		UPDATE incidents
-		   SET status = 'acknowledged', acknowledged_by = $1, acknowledged_at = $2
-		 WHERE id = $3 AND status = 'triggered'`,
-		userID, time.Now().Unix(), incidentID)
+		   SET status = 'acknowledged', acknowledged_by = $1, acknowledged_by_service_account_id = $2,
+		       acknowledged_at = $3
+		 WHERE id = $4 AND status = 'triggered'`,
+		userID, serviceAccountID, time.Now().Unix(), incidentID)
 	if err != nil {
 		return false, err
 	}
@@ -283,7 +328,7 @@ func acknowledgeIncident(ctx context.Context, q querier, incidentID, userID int6
 	if err := stopEscalation(ctx, q, incidentID); err != nil {
 		return false, err
 	}
-	return true, logEvent(ctx, q, incidentID, evAcknowledged, &userID, nil, nil)
+	return true, logEvent(ctx, q, incidentID, evAcknowledged, userID, serviceAccountID, nil, nil)
 }
 
 // openIncidentForAlert returns the open incident an alert currently belongs to,

internal/api/incidents.go

@@ -100,6 +100,10 @@ func handleListIncidents(db *sql.DB) http.HandlerFunc {
 			}
 			incidents = append(incidents, i)
 		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		respond(w, http.StatusOK, incidents)
 	}
 }
@@ -157,9 +161,15 @@ func handleIncidentTimeline(db *sql.DB) http.HandlerFunc {
 
 		rows, err := db.QueryContext(r.Context(), `
 			SELECT e.id, e.incident_id, e.type, e.user_id, u.username,
+			       e.service_account_id, sa.name,
+			       e.actor_user_id, au.username,
+			       e.actor_service_account_id, asa.name,
 			       e.alert_id, e.detail, e.created_at
 			FROM   incident_events e
 			LEFT   JOIN users u ON u.id = e.user_id
+			LEFT   JOIN service_accounts sa ON sa.id = e.service_account_id
+			LEFT   JOIN users au ON au.id = e.actor_user_id
+			LEFT   JOIN service_accounts asa ON asa.id = e.actor_service_account_id
 			WHERE  e.incident_id = $1
 			ORDER  BY e.created_at ASC, e.id ASC`, id)
 		if err != nil {
@@ -173,6 +183,9 @@ func handleIncidentTimeline(db *sql.DB) http.HandlerFunc {
 			var e models.IncidentEvent
 			var ts int64
 			if err := rows.Scan(&e.ID, &e.IncidentID, &e.Type, &e.UserID, &e.Username,
+				&e.ServiceAccountID, &e.ServiceAccountName,
+				&e.ActorUserID, &e.ActorUsername,
+				&e.ActorServiceAccountID, &e.ActorServiceAccountName,
 				&e.AlertID, &e.Detail, &ts); err != nil {
 				respond(w, http.StatusInternalServerError, errResp("internal error"))
 				return
@@ -180,6 +193,10 @@ func handleIncidentTimeline(db *sql.DB) http.HandlerFunc {
 			e.CreatedAt = time.Unix(ts, 0).UTC()
 			events = append(events, e)
 		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		respond(w, http.StatusOK, events)
 	}
 }
@@ -190,8 +207,8 @@ func handleIncidentAcknowledge(db *sql.DB) http.HandlerFunc {
 		if !ok {
 			return
 		}
-		user, _ := userFromContext(r.Context())
-		acked, err := acknowledgeIncident(r.Context(), db, id, user.ID)
+		userID, saID := callerActorIDs(r.Context())
+		acked, err := acknowledgeIncidentAs(r.Context(), db, id, userID, saID)
 		if err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
@@ -223,13 +240,14 @@ func handleIncidentUnacknowledge(db *sql.DB) http.HandlerFunc {
 		if !ok {
 			return
 		}
-		user, _ := userFromContext(r.Context())
+		userID, saID := callerActorIDs(r.Context())
 		if !updateOpenIncident(w, r, db, id,
-			`UPDATE incidents SET status = 'triggered', acknowledged_by = NULL, acknowledged_at = NULL
+			`UPDATE incidents SET status = 'triggered', acknowledged_by = NULL,
+			        acknowledged_by_service_account_id = NULL, acknowledged_at = NULL
 			  WHERE id = $1 AND resolved_at IS NULL`, id) {
 			return
 		}
-		if err := logEvent(r.Context(), db, id, evUnacknowledged, &user.ID, nil, nil); err != nil {
+		if err := logEvent(r.Context(), db, id, evUnacknowledged, userID, saID, nil, nil); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
@@ -247,7 +265,7 @@ func handleIncidentResolve(db *sql.DB) http.HandlerFunc {
 		if !ok {
 			return
 		}
-		user, _ := userFromContext(r.Context())
+		userID, saID := callerActorIDs(r.Context())
 		// The body is optional: clients that predate resolution notes send none.
 		var req struct {
 			Resolution string `json:"resolution"`
@@ -268,12 +286,12 @@ func handleIncidentResolve(db *sql.DB) http.HandlerFunc {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
-		if err := logEvent(r.Context(), db, id, evResolved, &user.ID, nil, nil); err != nil {
+		if err := logEvent(r.Context(), db, id, evResolved, userID, saID, nil, nil); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
 		if req.Resolution != "" {
-			if err := logEvent(r.Context(), db, id, evResolutionNote, &user.ID, nil, &req.Resolution); err != nil {
+			if err := logEvent(r.Context(), db, id, evResolutionNote, userID, saID, nil, &req.Resolution); err != nil {
 				respond(w, http.StatusInternalServerError, errResp("internal error"))
 				return
 			}
@@ -311,8 +329,10 @@ func handleIncidentAssign(db *sql.DB) http.HandlerFunc {
 			req.UserID, id) {
 			return
 		}
-		// On an "assigned" event user_id is the assignee, not the actor.
-		if err := logEvent(r.Context(), db, id, evAssigned, &req.UserID, nil, nil); err != nil {
+		// On an "assigned" event user_id is the assignee; the actor goes in
+		// the actor_* columns.
+		actorUserID, actorSAID := callerActorIDs(r.Context())
+		if err := logAssignedEvent(r.Context(), db, id, req.UserID, actorUserID, actorSAID); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
@@ -363,14 +383,14 @@ func handleIncidentSnooze(db *sql.DB) http.HandlerFunc {
 			return
 		}
 
-		user, _ := userFromContext(r.Context())
+		userID, saID := callerActorIDs(r.Context())
 		if !updateOpenIncident(w, r, db, id,
 			"UPDATE incidents SET snoozed_until = $1 WHERE id = $2 AND resolved_at IS NULL",
 			until.Unix(), id) {
 			return
 		}
 		detail := until.UTC().Format(time.RFC3339)
-		if err := logEvent(r.Context(), db, id, evSnoozed, &user.ID, nil, &detail); err != nil {
+		if err := logEvent(r.Context(), db, id, evSnoozed, userID, saID, nil, &detail); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
@@ -384,12 +404,12 @@ func handleIncidentUnsnooze(db *sql.DB) http.HandlerFunc {
 		if !ok {
 			return
 		}
-		user, _ := userFromContext(r.Context())
+		userID, saID := callerActorIDs(r.Context())
 		if !updateOpenIncident(w, r, db, id,
 			"UPDATE incidents SET snoozed_until = NULL WHERE id = $1 AND resolved_at IS NULL", id) {
 			return
 		}
-		if err := logEvent(r.Context(), db, id, evUnsnoozed, &user.ID, nil, nil); err != nil {
+		if err := logEvent(r.Context(), db, id, evUnsnoozed, userID, saID, nil, nil); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
@@ -413,6 +433,11 @@ func handleIncidentArchive(db *sql.DB) http.HandlerFunc {
 			respond(w, http.StatusNotFound, errResp("incident not found"))
 			return
 		}
+		userID, saID := callerActorIDs(r.Context())
+		if err := logEvent(r.Context(), db, id, evArchived, userID, saID, nil, nil); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		respondIncident(w, r, db, id)
 	}
 }
@@ -433,6 +458,11 @@ func handleIncidentUnarchive(db *sql.DB) http.HandlerFunc {
 			respond(w, http.StatusNotFound, errResp("incident not found"))
 			return
 		}
+		userID, saID := callerActorIDs(r.Context())
+		if err := logEvent(r.Context(), db, id, evUnarchived, userID, saID, nil, nil); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		w.WriteHeader(http.StatusNoContent)
 	}
 }
@@ -466,27 +496,32 @@ func handleCreateNote(db *sql.DB) http.HandlerFunc {
 			return
 		}
 
-		user, _ := userFromContext(r.Context())
+		caller, _ := callerFromContext(r.Context())
+		userID, saID := callerActorIDs(r.Context())
 		now := time.Now()
 		var eventID int64
 		err := db.QueryRowContext(r.Context(), `
-			INSERT INTO incident_events (incident_id, type, user_id, detail, created_at)
-			VALUES ($1, $2, $3, $4, $5)
-			RETURNING id`, id, noteType, user.ID, req.Content, now.Unix()).Scan(&eventID)
+			INSERT INTO incident_events (incident_id, type, user_id, service_account_id, detail, created_at)
+			VALUES ($1, $2, $3, $4, $5, $6)
+			RETURNING id`, id, noteType, userID, saID, req.Content, now.Unix()).Scan(&eventID)
 		if err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
 
-		respond(w, http.StatusCreated, models.IncidentEvent{
+		resp := models.IncidentEvent{
 			ID:         eventID,
 			IncidentID: id,
 			Type:       noteType,
-			UserID:     &user.ID,
-			Username:   &user.Username,
 			Detail:     &req.Content,
 			CreatedAt:  now.UTC().Truncate(time.Second),
-		})
+		}
+		if u, ok := caller.AsHuman(); ok {
+			resp.UserID, resp.Username = &u.ID, &u.Username
+		} else if saName, ok := caller.ServiceAccountName(); ok {
+			resp.ServiceAccountID, resp.ServiceAccountName = saID, &saName
+		}
+		respond(w, http.StatusCreated, resp)
 	}
 }
 
@@ -504,11 +539,12 @@ func handleDeleteNote(db *sql.DB) http.HandlerFunc {
 			return
 		}
 
-		user, _ := userFromContext(r.Context())
+		userID, saID := callerActorIDs(r.Context())
 		res, err := db.ExecContext(r.Context(), `
 			DELETE FROM incident_events
-			 WHERE id = $1 AND incident_id = $2 AND type IN ($3, $4) AND user_id = $5`,
-			eventID, id, evNote, evResolutionNote, user.ID)
+			 WHERE id = $1 AND incident_id = $2 AND type IN ($3, $4)
+			   AND (user_id = $5 OR service_account_id = $6)`,
+			eventID, id, evNote, evResolutionNote, userID, saID)
 		if err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return

internal/api/incidents_test.go

@@ -8,6 +8,7 @@ import (
 	"time"
 
 	"git.ryuvia.com/niklas/terdut-server/internal/api"
+	"git.ryuvia.com/niklas/terdut-server/internal/models"
 )
 
 // amAlert builds one alert of a webhook payload.
@@ -642,6 +643,106 @@ func TestIncident_ArchiveRoundTrip(t *testing.T) {
 	}
 }
 
+// ---------------------------------------------------------------------------
+// Service accounts (terdut-server#25)
+// ---------------------------------------------------------------------------
+
+// TestServiceAccount_CanActOnItsTeamsIncidents is #25's regression test.
+// Before the fix: acknowledge/resolve/snooze/create-note each 500'd (writing
+// acknowledged_by/user_id = 0, violating the users(id) FK for a service
+// account), and delete-note silently matched zero rows (WHERE user_id = 0)
+// instead of deleting.
+func TestServiceAccount_CanActOnItsTeamsIncidents(t *testing.T) {
+	s := newTS(t)
+	instanceKey := createServiceAccount(t, s, s.key, "operator", models.ServiceAccountScopeInstance, 0)
+	teamA := createTeamAs(t, s, instanceKey, "team-a")
+	keyA := createServiceAccount(t, s, instanceKey, "team-a-sa", models.ServiceAccountScopeTeam, teamA)
+
+	var integration struct {
+		Key string `json:"key"`
+	}
+	decode(t, s.reqAs(t, keyA, http.MethodPost, "/api/teams/"+id64(teamA)+"/integrations",
+		map[string]string{"name": "test"}), &integration)
+	postToIntegration(t, s, integration.Key, "fp-sa", "SAIncident") // incident 1
+
+	// Acknowledge.
+	resp := s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/acknowledge", nil)
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("service account acknowledge: %d", resp.StatusCode)
+	}
+	var inc map[string]any
+	decode(t, resp, &inc)
+	if inc["acknowledged_by_service_account_id"] == nil {
+		t.Error("expected acknowledged_by_service_account_id to be set")
+	}
+	if inc["acknowledged_by_id"] != nil {
+		t.Errorf("expected acknowledged_by_id to stay nil for a service-account actor, got %v", inc["acknowledged_by_id"])
+	}
+
+	// Unacknowledge.
+	resp = s.reqAs(t, keyA, http.MethodDelete, "/api/incidents/1/acknowledge", nil)
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusNoContent {
+		t.Errorf("service account unacknowledge: %d", resp.StatusCode)
+	}
+
+	// Snooze, then unsnooze.
+	resp = s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/snooze",
+		map[string]string{"duration": "1h"})
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		t.Errorf("service account snooze: %d", resp.StatusCode)
+	}
+	resp = s.reqAs(t, keyA, http.MethodDelete, "/api/incidents/1/snooze", nil)
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusNoContent {
+		t.Errorf("service account unsnooze: %d", resp.StatusCode)
+	}
+
+	// Create, then delete, a note.
+	var note map[string]any
+	decode(t, s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/notes",
+		map[string]string{"content": "looking into it"}), &note)
+	if note["service_account_id"] == nil {
+		t.Error("expected service_account_id on the note event")
+	}
+	if note["user_id"] != nil {
+		t.Errorf("expected no user_id on a service-account note, got %v", note["user_id"])
+	}
+	noteID := int(note["id"].(float64))
+	delResp := s.reqAs(t, keyA, http.MethodDelete, fmt.Sprintf("/api/incidents/1/notes/%d", noteID), nil)
+	delResp.Body.Close()
+	if delResp.StatusCode != http.StatusNoContent {
+		t.Errorf("service account deleting its own note: %d", delResp.StatusCode)
+	}
+
+	// Resolve.
+	resp = s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/resolve", nil)
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		t.Errorf("service account resolve: %d", resp.StatusCode)
+	}
+}
+
+// Regression guard: a human actor must still write only the human columns,
+// unaffected by the service-account branch added above.
+func TestIncident_AcknowledgeStillWritesOnlyHumanColumn(t *testing.T) {
+	s := newTS(t)
+	postWebhook(t, s, []map[string]any{
+		amAlert("fp-human-ack", "Z", "firing", "2026-05-20T10:00:00Z", zeroTime, nil),
+	})
+
+	var inc map[string]any
+	decode(t, s.req(t, http.MethodPost, "/api/incidents/1/acknowledge", nil), &inc)
+	if inc["acknowledged_by_id"] == nil {
+		t.Error("expected acknowledged_by_id to be set for a human actor")
+	}
+	if inc["acknowledged_by_service_account_id"] != nil {
+		t.Errorf("expected acknowledged_by_service_account_id to stay nil for a human actor, got %v",
+			inc["acknowledged_by_service_account_id"])
+	}
+}
+
 func TestSweeper_ArchivesResolvedIncidents(t *testing.T) {
 	s := newTS(t)
 	postWebhook(t, s, []map[string]any{
@@ -769,3 +870,105 @@ func contains(haystack []string, needle string) bool {
 	}
 	return false
 }
+
+// ---------------------------------------------------------------------------
+// Actor on assign / archive / unarchive (terdut-server#35)
+// ---------------------------------------------------------------------------
+
+// lastEvent returns the newest timeline event of the given type.
+func lastEvent(t *testing.T, events []map[string]any, typ string) map[string]any {
+	t.Helper()
+	for i := len(events) - 1; i >= 0; i-- {
+		if events[i]["type"] == typ {
+			return events[i]
+		}
+	}
+	t.Fatalf("no %q event in %v", typ, eventTypes(events))
+	return nil
+}
+
+func TestIncident_AssignRecordsHumanActor(t *testing.T) {
+	s := newTS(t)
+	postWebhook(t, s, []map[string]any{
+		amAlert("fp-asg-actor", "Assignable", "firing", "2026-05-20T10:00:00Z", zeroTime, nil),
+	})
+	s.req(t, http.MethodPost, "/api/users",
+		map[string]string{"username": "alice", "email": "alice@test.com"}).Body.Close()
+	s.req(t, http.MethodPost, "/api/incidents/1/assign", map[string]any{"user_id": 2}).Body.Close()
+
+	ev := lastEvent(t, timeline(t, s, 1), "assigned")
+	if ev["username"] != "alice" {
+		t.Errorf("expected the assignee alice in username, got %v", ev["username"])
+	}
+	if ev["actor_user_id"] == nil || ev["actor_username"] == nil {
+		t.Errorf("expected the assigning human in actor_*, got %v", ev)
+	}
+	if ev["actor_service_account_id"] != nil {
+		t.Errorf("expected no service-account actor, got %v", ev["actor_service_account_id"])
+	}
+}
+
+func TestIncident_ArchiveUnarchiveRecordHumanActor(t *testing.T) {
+	s := newTS(t)
+	postWebhook(t, s, []map[string]any{
+		amAlert("fp-arc-actor", "Archivable", "firing", "2026-05-20T10:00:00Z", zeroTime, nil),
+	})
+	s.req(t, http.MethodPost, "/api/incidents/1/resolve", nil).Body.Close()
+	s.req(t, http.MethodPost, "/api/incidents/1/archive", nil).Body.Close()
+	s.req(t, http.MethodDelete, "/api/incidents/1/archive", nil).Body.Close()
+
+	events := timeline(t, s, 1)
+	for _, typ := range []string{"archived", "unarchived"} {
+		ev := lastEvent(t, events, typ)
+		if ev["user_id"] == nil || ev["service_account_id"] != nil {
+			t.Errorf("%s: expected only the human actor, got %v", typ, ev)
+		}
+	}
+}
+
+func TestServiceAccount_AssignArchiveUnarchiveRecordActor(t *testing.T) {
+	s := newTS(t)
+	instanceKey := createServiceAccount(t, s, s.key, "operator", models.ServiceAccountScopeInstance, 0)
+	teamA := createTeamAs(t, s, instanceKey, "team-a")
+	keyA := createServiceAccount(t, s, instanceKey, "team-a-sa", models.ServiceAccountScopeTeam, teamA)
+	var integration struct {
+		Key string `json:"key"`
+	}
+	decode(t, s.reqAs(t, keyA, http.MethodPost, "/api/teams/"+id64(teamA)+"/integrations",
+		map[string]string{"name": "test"}), &integration)
+	postToIntegration(t, s, integration.Key, "fp-sa-35", "SA35") // incident 1
+
+	resp := s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/assign", map[string]any{"user_id": 1})
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("service account assign: %d", resp.StatusCode)
+	}
+	resp = s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/resolve", nil)
+	resp.Body.Close()
+	resp = s.reqAs(t, keyA, http.MethodPost, "/api/incidents/1/archive", nil)
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusOK {
+		t.Fatalf("service account archive: %d", resp.StatusCode)
+	}
+	resp = s.reqAs(t, keyA, http.MethodDelete, "/api/incidents/1/archive", nil)
+	resp.Body.Close()
+	if resp.StatusCode != http.StatusNoContent {
+		t.Fatalf("service account unarchive: %d", resp.StatusCode)
+	}
+
+	var events []map[string]any
+	decode(t, s.reqAs(t, keyA, http.MethodGet, "/api/incidents/1/timeline", nil), &events)
+	asg := lastEvent(t, events, "assigned")
+	if asg["actor_service_account_id"] == nil || asg["actor_user_id"] != nil {
+		t.Errorf("assigned: expected only the service-account actor, got %v", asg)
+	}
+	if asg["user_id"] == nil {
+		t.Errorf("assigned: user_id must stay the assignee, got %v", asg)
+	}
+	for _, typ := range []string{"archived", "unarchived"} {
+		ev := lastEvent(t, events, typ)
+		if ev["service_account_id"] == nil || ev["user_id"] != nil {
+			t.Errorf("%s: expected only the service-account actor, got %v", typ, ev)
+		}
+	}
+}

internal/api/middleware.go

@@ -79,6 +79,28 @@ func AuthMiddleware(db *sql.DB) func(http.Handler) http.Handler {
 	}
 }
 
+// securityHeaders sets headers that cost nothing to send on every response,
+// API or static site alike. nosniff is unconditional; HSTS only fires once
+// cookieSecure's signal says the browser is actually looking at this server
+// over HTTPS — TLS terminates at the gateway, which (as of this writing) sets
+// neither header itself.
+//
+// max-age is 180 days rather than the usual year-plus: short enough that if
+// HTTPS here ever broke for real, the header would age out of a browser's
+// cache well within a release cycle instead of locking anyone out of a
+// working server. Raise it once this has run clean for a while.
+func securityHeaders(publicURL string) func(http.Handler) http.Handler {
+	return func(next http.Handler) http.Handler {
+		return http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+			w.Header().Set("X-Content-Type-Options", "nosniff")
+			if cookieSecure(publicURL, r) {
+				w.Header().Set("Strict-Transport-Security", "max-age=15552000; includeSubDomains")
+			}
+			next.ServeHTTP(w, r)
+		})
+	}
+}
+
 // AdminOnly rejects a caller who is not a system administrator. It runs inside
 // AuthMiddleware's group, so by the time it sees a request the caller is known.
 //
@@ -113,10 +135,16 @@ func requireSelfOrAdmin(w http.ResponseWriter, r *http.Request, targetID int64)
 }
 
 // apiKeyUser resolves an API key to its user and stamps its last use.
+// expires_at IS NULL OR > now is part of the lookup itself, the same way
+// serveAs's disabled_at check is: an expired key is one that cannot
+// authenticate, by construction, rather than one that happens to still
+// resolve and has to be caught afterwards.
 func apiKeyUser(ctx context.Context, db *sql.DB, token string) (int64, bool) {
 	var keyID, userID int64
 	err := db.QueryRowContext(ctx,
-		"SELECT id, user_id FROM api_keys WHERE key_hash = $1", hashToken(token),
+		`SELECT id, user_id FROM api_keys
+		  WHERE key_hash = $1 AND (expires_at IS NULL OR expires_at > $2)`,
+		hashToken(token), time.Now().Unix(),
 	).Scan(&keyID, &userID)
 	if err != nil {
 		return 0, false

internal/api/notifier.go

@@ -38,6 +38,13 @@ const (
 	ackTokenTTL = 24 * time.Hour
 )
 
+// notifierLockKey is the Postgres advisory lock the notifier takes for the
+// duration of each pass, so that running more than one replica does not
+// deliver (or double-deliver) the same notification from more than one of
+// them at once. Its value has no meaning beyond being distinct from
+// archiverLockKey.
+const notifierLockKey int64 = 7265_0002
+
 // Notification kinds, recording why a push was sent.
 const (
 	notifyTriggered = "triggered"
@@ -97,6 +104,10 @@ var notifyClient = &http.Client{Timeout: 10 * time.Second}
 
 // StartNotifier delivers queued notifications until ctx is cancelled, starting
 // with an immediate pass so a restart flushes whatever the last one left behind.
+//
+// Each pass runs under notifierLockKey (see withAdvisoryLock), so that on more
+// than one replica only whichever instance's tick takes the lock first actually
+// delivers; the rest skip that tick rather than racing the same pass.
 func StartNotifier(ctx context.Context, db *sql.DB, cfg NotifyConfig) {
 	if !cfg.enabled() {
 		log.Print("notifier: disabled (no ntfy URL configured)")
@@ -107,11 +118,17 @@ func StartNotifier(ctx context.Context, db *sql.DB, cfg NotifyConfig) {
 	ticker := time.NewTicker(notifyInterval)
 	defer ticker.Stop()
 
-	NotifySweep(ctx, db, cfg)
+	sweep := func() {
+		withAdvisoryLock(ctx, db, notifierLockKey, "notifier", func() {
+			NotifySweep(ctx, db, cfg)
+		})
+	}
+
+	sweep()
 	for {
 		select {
 		case <-ticker.C:
-			NotifySweep(ctx, db, cfg)
+			sweep()
 		case <-ctx.Done():
 			return
 		}
@@ -240,7 +257,7 @@ func deliverPending(ctx context.Context, db *sql.DB, cfg NotifyConfig) {
 		}
 		// Logged, not returned: the page has already gone out, and treating a
 		// failed timeline write as a failed delivery would send it again.
-		if err := logEvent(ctx, db, n.incidentID, eventNotified, n.userID, nil, &n.kind); err != nil {
+		if err := logEvent(ctx, db, n.incidentID, eventNotified, n.userID, nil, nil, &n.kind); err != nil {
 			log.Printf("notifier: log delivery of %d: %v", n.id, err)
 		}
 		sent++
@@ -295,7 +312,7 @@ func markFailed(ctx context.Context, db *sql.DB, n outboxRow, cause error) {
 		return
 	}
 	detail := fmt.Sprintf("%s: %s", n.kind, cause)
-	if err := logEvent(ctx, db, n.incidentID, eventNotifyFailed, n.userID, nil, &detail); err != nil {
+	if err := logEvent(ctx, db, n.incidentID, eventNotifyFailed, n.userID, nil, nil, &detail); err != nil {
 		log.Printf("notifier: log failure of %d: %v", n.id, err)
 	}
 }

internal/api/oidc.go

@@ -121,12 +121,12 @@ func ssoRedirect(w http.ResponseWriter, r *http.Request, code ssoError) {
 func handleOIDCLogin(db *sql.DB, prov *oidc.Provider, limiter *loginLimiter, publicURL string) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		addrKey := "oidc:" + clientAddr(r)
-		if limiter.blocked(addrKey, oidcStartMaxPerAddr) {
+		if limiter.blocked(r.Context(), addrKey, oidcStartMaxPerAddr) {
 			w.Header().Set("Retry-After", strconv.Itoa(int(loginWindow.Seconds())))
 			respond(w, http.StatusTooManyRequests, errResp("too many sign-in attempts, try again later"))
 			return
 		}
-		limiter.fail(addrKey)
+		limiter.fail(r.Context(), addrKey)
 
 		state, stateHash, err := randomToken()
 		if err != nil {
@@ -160,6 +160,9 @@ func handleOIDCLogin(db *sql.DB, prov *oidc.Provider, limiter *loginLimiter, pub
 			return
 		}
 
+		// #nosec G124 -- HttpOnly/SameSite are literal below; Secure is
+		// cookieSecure(publicURL, r), not a literal true, which is what
+		// trips this rule. See cookieSecure's own doc comment in auth.go.
 		http.SetCookie(w, &http.Cookie{
 			Name:     oidcStateCookie,
 			Value:    state,
@@ -182,6 +185,9 @@ func handleOIDCCallback(db *sql.DB, prov *oidc.Provider, publicURL string) http.
 	return func(w http.ResponseWriter, r *http.Request) {
 		// The state cookie has done its job once the callback arrives, whatever
 		// the outcome.
+		// #nosec G124 -- HttpOnly/SameSite are literal below; Secure is
+		// cookieSecure(publicURL, r), not a literal true, which is what
+		// trips this rule. See cookieSecure's own doc comment in auth.go.
 		http.SetCookie(w, &http.Cookie{
 			Name: oidcStateCookie, Value: "", Path: "/api/oidc", MaxAge: -1,
 			HttpOnly: true, Secure: cookieSecure(publicURL, r), SameSite: http.SameSiteLaxMode,
@@ -189,7 +195,13 @@ func handleOIDCCallback(db *sql.DB, prov *oidc.Provider, publicURL string) http.
 
 		q := r.URL.Query()
 		if e := q.Get("error"); e != "" {
-			log.Printf("oidc: provider returned error %q: %s", e, q.Get("error_description"))
+			// %q on both: this runs before state is checked against the
+			// cookie, so error and error_description are still whatever the
+			// request's query string says, not yet known to be the real
+			// provider's. %q keeps a crafted value (say, one holding a
+			// newline) from forging a second log line rather than just
+			// being a quoted string within this one.
+			log.Printf("oidc: provider returned error %q: %q", e, q.Get("error_description")) // #nosec G706 -- both %q
 			ssoRedirect(w, r, ssoDenied)
 			return
 		}
@@ -226,7 +238,7 @@ func handleOIDCCallback(db *sql.DB, prov *oidc.Provider, publicURL string) http.
 
 		grants := oidc.ComputeGrants(cfg, identity.Groups)
 		if !grants.Admitted {
-			log.Printf("oidc: %q (%s) is in none of the allowed groups", identity.Username, identity.Subject)
+			log.Printf("oidc: %q (%q) is in none of the allowed groups", identity.Username, identity.Subject) // #nosec G706 -- both %q
 			ssoRedirect(w, r, ssoNotAllowed)
 			return
 		}
@@ -243,11 +255,11 @@ func handleOIDCCallback(db *sql.DB, prov *oidc.Provider, publicURL string) http.
 		if err != nil {
 			var se ssoError
 			if errors.As(err, &se) {
-				log.Printf("oidc: refused %q (%s): %v", identity.Username, identity.Subject, se)
+				log.Printf("oidc: refused %q (%q): %v", identity.Username, identity.Subject, se) // #nosec G706 -- both %q
 				ssoRedirect(w, r, se)
 				return
 			}
-			log.Printf("oidc: sign in %q: %v", identity.Username, err)
+			log.Printf("oidc: sign in %q: %v", identity.Username, err) // #nosec G706 -- %q
 			ssoRedirect(w, r, ssoFailed)
 			return
 		}

internal/api/rate_limiter_test.go

@@ -0,0 +1,161 @@
+package api
+
+// This file is internal (package api, not api_test) because loginLimiter and
+// its blocked/fail/clear methods are unexported, and TestLoginLimiter_SharedAcrossReplicas
+// specifically needs to construct two separate loginLimiter values pointed at
+// one database — standing in for two replicas — which only this package can
+// do. It duplicates testdb_test.go's newTestDB/withSearchPath rather than
+// importing them: those live in the separate api_test package, compiled from
+// this directory's external test files, and are not visible here. Same
+// reasoning as advisory_lock_test.go, which makes the same trade for the
+// same reason.
+
+import (
+	"context"
+	"database/sql"
+	"fmt"
+	"net/url"
+	"os"
+	"strings"
+	"testing"
+
+	"git.ryuvia.com/niklas/terdut-server/internal/db"
+	_ "github.com/jackc/pgx/v5/stdlib"
+)
+
+var rateLimiterSchemaSeq int
+
+// rateLimiterTestDB returns a migrated database private to this test.
+func rateLimiterTestDB(t *testing.T) *sql.DB {
+	t.Helper()
+
+	dsn := os.Getenv("TERDUT_TEST_DSN")
+	if dsn == "" {
+		t.Fatalf("TERDUT_TEST_DSN is not set: these tests need Postgres.\n" +
+			"Run `make test-db` for a local one, then\n" +
+			"  export TERDUT_TEST_DSN=postgres://terdut:terdut@localhost:5432/terdut_test?sslmode=disable")
+	}
+
+	rateLimiterSchemaSeq++
+	schema := fmt.Sprintf("test_rl_%d_%d", os.Getpid(), rateLimiterSchemaSeq)
+
+	admin, err := sql.Open("pgx", dsn)
+	if err != nil {
+		t.Fatalf("connect to TERDUT_TEST_DSN: %v", err)
+	}
+	defer admin.Close()
+	if _, err := admin.Exec("CREATE SCHEMA " + schema); err != nil {
+		t.Fatalf("create schema %s: %v", schema, err)
+	}
+
+	database, err := db.Open(rateLimiterWithSearchPath(dsn, schema))
+	if err != nil {
+		t.Fatalf("open db: %v", err)
+	}
+	if err := db.Migrate(database); err != nil {
+		t.Fatalf("migrate: %v", err)
+	}
+
+	t.Cleanup(func() {
+		database.Close()
+		cleanup, err := sql.Open("pgx", dsn)
+		if err != nil {
+			return
+		}
+		defer cleanup.Close()
+		if _, err := cleanup.Exec("DROP SCHEMA " + schema + " CASCADE"); err != nil {
+			t.Logf("drop schema %s: %v", schema, err)
+		}
+	})
+
+	return database
+}
+
+func rateLimiterWithSearchPath(dsn, schema string) string {
+	opt := "-csearch_path=" + schema
+	if strings.HasPrefix(dsn, "postgres://") || strings.HasPrefix(dsn, "postgresql://") {
+		u, err := url.Parse(dsn)
+		if err == nil {
+			q := u.Query()
+			q.Set("options", opt)
+			u.RawQuery = q.Encode()
+			return u.String()
+		}
+	}
+	return dsn + " options='" + opt + "'"
+}
+
+func TestLoginLimiter_BlocksAtMax(t *testing.T) {
+	database := rateLimiterTestDB(t)
+	ctx := context.Background()
+	l := newLoginLimiter(database)
+
+	for range 3 {
+		if l.blocked(ctx, "k", 3) {
+			t.Fatal("blocked before reaching max")
+		}
+		l.fail(ctx, "k")
+	}
+	if !l.blocked(ctx, "k", 3) {
+		t.Fatal("not blocked after reaching max")
+	}
+}
+
+func TestLoginLimiter_ClearResetsTheCount(t *testing.T) {
+	database := rateLimiterTestDB(t)
+	ctx := context.Background()
+	l := newLoginLimiter(database)
+
+	l.fail(ctx, "k")
+	l.fail(ctx, "k")
+	l.clear(ctx, "k")
+
+	if l.blocked(ctx, "k", 1) {
+		t.Fatal("still blocked after clear")
+	}
+}
+
+func TestLoginLimiter_KeysAreIndependent(t *testing.T) {
+	database := rateLimiterTestDB(t)
+	ctx := context.Background()
+	l := newLoginLimiter(database)
+
+	l.fail(ctx, "a")
+	if l.blocked(ctx, "b", 1) {
+		t.Fatal("failing one key blocked an unrelated one")
+	}
+}
+
+// TestLoginLimiter_SharedAcrossReplicas is the regression test for the gap
+// this migration closes: an in-memory limiter would let each replica count
+// independently, so a caller hitting two different pods could rack up
+// max*replicaCount failures before either one blocked. Two loginLimiter
+// values sharing one database, standing in for two replicas behind the same
+// load balancer, must instead see one combined count.
+func TestLoginLimiter_SharedAcrossReplicas(t *testing.T) {
+	database := rateLimiterTestDB(t)
+	ctx := context.Background()
+	replicaA := newLoginLimiter(database)
+	replicaB := newLoginLimiter(database)
+
+	const max = 4
+	// Alternate which "replica" records the failure, as a real deployment
+	// would split requests across pods.
+	for i := range max {
+		replica := replicaA
+		if i%2 == 1 {
+			replica = replicaB
+		}
+		if replicaA.blocked(ctx, "k", max) || replicaB.blocked(ctx, "k", max) {
+			t.Fatalf("blocked after only %d of %d failures", i, max)
+		}
+		replica.fail(ctx, "k")
+	}
+
+	if !replicaA.blocked(ctx, "k", max) {
+		t.Fatal("replica A does not see the combined count as blocked")
+	}
+	if !replicaB.blocked(ctx, "k", max) {
+		t.Fatal("replica B does not see the combined count as blocked")
+	}
+}

internal/api/router.go

@@ -23,13 +23,14 @@ func NewRouter(db *sql.DB, notify NotifyConfig, cfg config.Config, version strin
 	// One limiter each, both process-wide for the life of the router: login
 	// counts failed passwords, sign-up counts account creation, and mixing the
 	// two would let a burst of sign-ups lock somebody out of logging in.
-	loginLimit := newLoginLimiter()
-	signupLimiter := newLoginLimiter()
-	oidcLimit := newLoginLimiter()
+	loginLimit := newLoginLimiter(db)
+	signupLimiter := newLoginLimiter(db)
+	oidcLimit := newLoginLimiter(db)
 
 	r := chi.NewRouter()
 	r.Use(middleware.Logger)
 	r.Use(middleware.Recoverer)
+	r.Use(securityHeaders(notify.PublicURL))
 
 	r.Get("/healthz", func(w http.ResponseWriter, r *http.Request) {
 		respond(w, http.StatusOK, map[string]string{"status": "ok"})
@@ -111,6 +112,7 @@ func NewRouter(db *sql.DB, notify NotifyConfig, cfg config.Config, version strin
 		r.Get("/api/users/{id}/teams", handleUserTeams(db))
 		r.Put("/api/users/{id}/notify", handleSetNotifyTarget(db))
 		r.Put("/api/users/{id}/password", handleSetPassword(db))
+		r.Get("/api/users/{id}/api-keys", handleListAPIKeys(db))
 		r.Post("/api/users/{id}/api-keys", handleCreateAPIKey(db))
 		r.Delete("/api/users/{id}/api-keys/{keyID}", handleDeleteAPIKey(db))
 

internal/api/schedule.go

@@ -235,6 +235,9 @@ func scheduleRange(ctx context.Context, db *sql.DB, teamID int64, from, to strin
 
 	clause := strings.Join(where, " AND ")
 
+	// #nosec G202 -- clause is built from sqlArgs.add's "$N" placeholders
+	// only, never a value; every value travels through args.all() as a
+	// bound parameter. See the sqlArgs doc comment in helpers.go.
 	rows, err := db.QueryContext(ctx, `
 		SELECT s.id, s.team_id, t.name, s.user_id, u.username, s.date, s.created_at
 		FROM   schedule_entries s

internal/api/security_headers_test.go

@@ -0,0 +1,115 @@
+package api_test
+
+import (
+	"bytes"
+	"net/http"
+	"strings"
+	"testing"
+
+	"git.ryuvia.com/niklas/terdut-server/internal/api"
+)
+
+// ---------------------------------------------------------------------------
+// Security headers
+// ---------------------------------------------------------------------------
+
+func TestSecurityHeaders_NosniffAlwaysSet(t *testing.T) {
+	s := newTS(t) // no PublicURL: the HTTPS signal is off
+	resp := s.req(t, http.MethodGet, "/api/me", nil)
+	defer resp.Body.Close()
+
+	if got := resp.Header.Get("X-Content-Type-Options"); got != "nosniff" {
+		t.Errorf("X-Content-Type-Options = %q, want nosniff", got)
+	}
+	if got := resp.Header.Get("Strict-Transport-Security"); got != "" {
+		t.Errorf("Strict-Transport-Security = %q, want unset without an https PublicURL", got)
+	}
+}
+
+func TestSecurityHeaders_HSTSWhenPublicURLIsHTTPS(t *testing.T) {
+	s := newTS(t, api.NotifyConfig{PublicURL: "https://terdut.example.com"})
+	resp := s.req(t, http.MethodGet, "/api/me", nil)
+	defer resp.Body.Close()
+
+	got := resp.Header.Get("Strict-Transport-Security")
+	if !strings.HasPrefix(got, "max-age=") || !strings.Contains(got, "includeSubDomains") {
+		t.Errorf("Strict-Transport-Security = %q, want a max-age with includeSubDomains", got)
+	}
+}
+
+// ---------------------------------------------------------------------------
+// Request body size limits
+// ---------------------------------------------------------------------------
+
+// TestBodySizeLimit_OrdinaryEndpointRejectsOversizedBody confirms an
+// unauthenticated endpoint can't be made to buffer an arbitrarily large body:
+// past maxBodyBytes, decodeJSON fails exactly as it would on any other
+// malformed body, rather than the server reading the whole thing first.
+func TestBodySizeLimit_OrdinaryEndpointRejectsOversizedBody(t *testing.T) {
+	s := newTS(t)
+
+	huge := bytes.Repeat([]byte("a"), 2<<20) // 2 MiB, past the 1 MiB default
+	body := []byte(`{"username":"` + string(huge) + `","password":"x"}`)
+
+	resp, err := http.Post(s.URL+"/api/login", "application/json", bytes.NewReader(body))
+	if err != nil {
+		t.Fatalf("POST /api/login: %v", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusBadRequest {
+		t.Errorf("status = %d, want %d (oversized body treated as invalid)", resp.StatusCode, http.StatusBadRequest)
+	}
+}
+
+// TestBodySizeLimit_WebhookAllowsLargerBodyThanDefault confirms the
+// Alertmanager webhook's separate, larger cap actually takes effect: a body
+// bigger than the ordinary default but within maxWebhookBodyBytes is still
+// accepted, not rejected by the smaller limit every other endpoint gets.
+func TestBodySizeLimit_WebhookAllowsLargerBodyThanDefault(t *testing.T) {
+	s := newTS(t)
+
+	// Padding kept inside one alert's annotation, comfortably past the 1 MiB
+	// default and still well under the webhook's 8 MiB cap.
+	padding := strings.Repeat("a", 3<<20) // 3 MiB
+	payload := `{"version":"4","status":"firing","groupKey":"big-group",` +
+		`"groupLabels":{"alertname":"BigAlert"},"alerts":[{"status":"firing",` +
+		`"labels":{"alertname":"BigAlert"},"annotations":{"note":"` + padding + `"},` +
+		`"startsAt":"2026-05-20T10:00:00Z","endsAt":"0001-01-01T00:00:00Z",` +
+		`"fingerprint":"fp-big"}]}`
+
+	resp, err := http.Post(s.URL+"/api/integrations/"+s.ingestKey+"/alertmanager",
+		"application/json", strings.NewReader(payload))
+	if err != nil {
+		t.Fatalf("POST webhook: %v", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusOK {
+		t.Errorf("status = %d, want %d (body under the webhook's own cap)", resp.StatusCode, http.StatusOK)
+	}
+}
+
+// TestBodySizeLimit_WebhookRejectsPastItsOwnCap confirms the webhook's larger
+// cap is still a cap, not an exemption from one.
+func TestBodySizeLimit_WebhookRejectsPastItsOwnCap(t *testing.T) {
+	s := newTS(t)
+
+	huge := strings.Repeat("a", 9<<20) // 9 MiB, past the 8 MiB webhook cap
+	payload := `{"version":"4","status":"firing","groupKey":"huge-group",` +
+		`"groupLabels":{"alertname":"HugeAlert"},"alerts":[{"status":"firing",` +
+		`"labels":{"alertname":"HugeAlert"},"annotations":{"note":"` + huge + `"},` +
+		`"startsAt":"2026-05-20T10:00:00Z","endsAt":"0001-01-01T00:00:00Z",` +
+		`"fingerprint":"fp-huge"}]}`
+
+	resp, err := http.Post(s.URL+"/api/integrations/"+s.ingestKey+"/alertmanager",
+		"application/json", strings.NewReader(payload))
+	if err != nil {
+		t.Fatalf("POST webhook: %v", err)
+	}
+	defer resp.Body.Close()
+
+	if resp.StatusCode != http.StatusBadRequest {
+		t.Errorf("status = %d, want %d (body past the webhook's own cap)", resp.StatusCode, http.StatusBadRequest)
+	}
+}

internal/api/signup.go

@@ -112,7 +112,7 @@ var errInviteUnusable = errors.New("invite is not usable")
 func handleSignup(db *sql.DB, limiter *loginLimiter, publicURL string) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		addr := clientAddr(r)
-		if limiter.blocked("signup:"+addr, maxSignupsPerAddr) {
+		if limiter.blocked(r.Context(), "signup:"+addr, maxSignupsPerAddr) {
 			respond(w, http.StatusTooManyRequests, errResp("too many sign-ups from this address"))
 			return
 		}
@@ -148,7 +148,7 @@ func handleSignup(db *sql.DB, limiter *loginLimiter, publicURL string) http.Hand
 			var err error
 			inv, err = loadInvite(r.Context(), db, req.Invite)
 			if err != nil {
-				limiter.fail("signup:" + addr)
+				limiter.fail(r.Context(), "signup:"+addr)
 				respond(w, http.StatusForbidden, errResp("this invite link is not usable"))
 				return
 			}

internal/api/stats.go

@@ -73,6 +73,10 @@ func handleStatsTop(db *sql.DB) http.HandlerFunc {
 			}
 			result = append(result, e)
 		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		respond(w, http.StatusOK, result)
 	}
 }
@@ -104,6 +108,10 @@ func handleStatsByHour(db *sql.DB) http.HandlerFunc {
 			}
 			counts[hr] = cnt
 		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 
 		type entry struct {
 			Hour  int   `json:"hour"`
@@ -146,6 +154,10 @@ func handleStatsByDay(db *sql.DB) http.HandlerFunc {
 			}
 			counts[dow] = cnt
 		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 
 		dayNames := [7]string{"Sunday", "Monday", "Tuesday", "Wednesday", "Thursday", "Friday", "Saturday"}
 		type entry struct {

internal/api/users.go

@@ -116,6 +116,10 @@ func handleListUsers(db *sql.DB) http.HandlerFunc {
 			u.DisabledAt = unixPtr(disabled)
 			users = append(users, u)
 		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
 		respond(w, http.StatusOK, users)
 	}
 }
@@ -234,6 +238,11 @@ func handleDeleteUser(db *sql.DB) http.HandlerFunc {
 	}
 }
 
+// maxAPIKeyExpiryDays bounds expires_in_days: generous enough for any real
+// rotation policy, tight enough to reject a typo (a year in hours, say) that
+// would otherwise mint a key that outlives the server by decades.
+const maxAPIKeyExpiryDays = 3650 // ~10 years
+
 func handleCreateAPIKey(db *sql.DB) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		userID, err := strconv.ParseInt(chi.URLParam(r, "id"), 10, 64)
@@ -247,6 +256,11 @@ func handleCreateAPIKey(db *sql.DB) http.HandlerFunc {
 
 		var req struct {
 			Name string `json:"name"`
+			// ExpiresInDays is optional and, left zero, means the key never
+			// expires — the only behavior any key had before this field
+			// existed, so an existing integration that does not send it is
+			// unaffected.
+			ExpiresInDays int64 `json:"expires_in_days,omitempty"`
 		}
 		if err := decodeJSON(r, &req); err != nil {
 			respond(w, http.StatusBadRequest, errResp("invalid request body"))
@@ -256,6 +270,10 @@ func handleCreateAPIKey(db *sql.DB) http.HandlerFunc {
 			respond(w, http.StatusBadRequest, errResp("name is required"))
 			return
 		}
+		if req.ExpiresInDays < 0 || req.ExpiresInDays > maxAPIKeyExpiryDays {
+			respond(w, http.StatusBadRequest, errResp("expires_in_days must be 0 (never expires) or up to "+strconv.Itoa(maxAPIKeyExpiryDays)))
+			return
+		}
 
 		var exists int
 		if err := db.QueryRowContext(r.Context(), "SELECT 1 FROM users WHERE id = $1", userID).Scan(&exists); err != nil {
@@ -268,18 +286,77 @@ func handleCreateAPIKey(db *sql.DB) http.HandlerFunc {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
+		var expiresAt *int64
+		var expiresAtTime *time.Time
+		if req.ExpiresInDays > 0 {
+			t := time.Now().AddDate(0, 0, int(req.ExpiresInDays)).UTC()
+			u := t.Unix()
+			expiresAt = &u
+			expiresAtTime = &t
+		}
 		var keyID int64
 		if err := db.QueryRowContext(r.Context(),
-			"INSERT INTO api_keys (user_id, key_hash, name) VALUES ($1, $2, $3) RETURNING id",
-			userID, hash, req.Name).Scan(&keyID); err != nil {
+			"INSERT INTO api_keys (user_id, key_hash, name, expires_at) VALUES ($1, $2, $3, $4) RETURNING id",
+			userID, hash, req.Name, expiresAt).Scan(&keyID); err != nil {
 			respond(w, http.StatusInternalServerError, errResp("internal error"))
 			return
 		}
-		key := models.APIKey{ID: keyID, UserID: userID, Name: req.Name, Key: raw, CreatedAt: time.Now().UTC()}
+		key := models.APIKey{
+			ID: keyID, UserID: userID, Name: req.Name, Key: raw,
+			CreatedAt: time.Now().UTC(), ExpiresAt: expiresAtTime,
+		}
 		respond(w, http.StatusCreated, key)
 	}
 }
 
+// handleListAPIKeys lists a user's own API keys: never the raw key itself
+// (only ever returned once, at creation), just enough to tell them apart,
+// see which are stale (last_used_at) and which are about to stop working
+// (expires_at) — the data handleCreateAPIKey and apiKeyUser's last-use stamp
+// already produce, with no endpoint to read it back until now.
+func handleListAPIKeys(db *sql.DB) http.HandlerFunc {
+	return func(w http.ResponseWriter, r *http.Request) {
+		userID, err := strconv.ParseInt(chi.URLParam(r, "id"), 10, 64)
+		if err != nil {
+			respond(w, http.StatusBadRequest, errResp("invalid user id"))
+			return
+		}
+		if !requireSelfOrAdmin(w, r, userID) {
+			return
+		}
+
+		rows, err := db.QueryContext(r.Context(),
+			`SELECT id, name, created_at, last_used_at, expires_at
+			   FROM api_keys WHERE user_id = $1 ORDER BY created_at DESC`, userID)
+		if err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
+		defer rows.Close()
+
+		keys := []models.APIKey{}
+		for rows.Next() {
+			var k models.APIKey
+			var created int64
+			var lastUsed, expires *int64
+			if err := rows.Scan(&k.ID, &k.Name, &created, &lastUsed, &expires); err != nil {
+				respond(w, http.StatusInternalServerError, errResp("internal error"))
+				return
+			}
+			k.UserID = userID
+			k.CreatedAt = time.Unix(created, 0).UTC()
+			k.LastUsedAt = unixPtr(lastUsed)
+			k.ExpiresAt = unixPtr(expires)
+			keys = append(keys, k)
+		}
+		if err := rows.Err(); err != nil {
+			respond(w, http.StatusInternalServerError, errResp("internal error"))
+			return
+		}
+		respond(w, http.StatusOK, keys)
+	}
+}
+
 func handleDeleteAPIKey(db *sql.DB) http.HandlerFunc {
 	return func(w http.ResponseWriter, r *http.Request) {
 		userID, err := strconv.ParseInt(chi.URLParam(r, "id"), 10, 64)

internal/db/db.go

@@ -1,6 +1,7 @@
 package db
 
 import (
+	"context"
 	"database/sql"
 	"embed"
 	"fmt"
@@ -62,6 +63,16 @@ func Open(dsn string) (*sql.DB, error) {
 	}
 }
 
+// migrationLockKey is the Postgres advisory lock Migrate holds for its whole
+// run. Two replicas starting at once would otherwise race the check-then-apply
+// loop below against schema_migrations: the loser could crash on a
+// duplicate-key insert, or contend with the winner's uncommitted DDL. Blocking
+// (pg_advisory_lock, not pg_try_advisory_lock as the archiver and notifier
+// use): on boot there is no later tick to defer to, so the right behaviour is
+// to wait for the other replica to finish migrating, not to skip ahead and
+// start serving against an unmigrated schema.
+const migrationLockKey int64 = 7265_0003
+
 // Migrate applies every embedded migration that has not been applied yet, in
 // filename order, recording each in schema_migrations.
 //
@@ -69,6 +80,22 @@ func Open(dsn string) (*sql.DB, error) {
 // migration that failed half way used to leave the schema in whatever state it
 // had reached. Postgres has transactional DDL, so the rollback is real.
 func Migrate(db *sql.DB) error {
+	ctx := context.Background()
+	conn, err := db.Conn(ctx)
+	if err != nil {
+		return fmt.Errorf("migrate: acquire connection: %w", err)
+	}
+	defer conn.Close()
+
+	if _, err := conn.ExecContext(ctx, "SELECT pg_advisory_lock($1)", migrationLockKey); err != nil {
+		return fmt.Errorf("migrate: acquire advisory lock: %w", err)
+	}
+	defer func() {
+		if _, err := conn.ExecContext(ctx, "SELECT pg_advisory_unlock($1)", migrationLockKey); err != nil {
+			log.Printf("migrate: release advisory lock: %v", err)
+		}
+	}()
+
 	if _, err := db.Exec(`CREATE TABLE IF NOT EXISTS schema_migrations (
 		version    TEXT PRIMARY KEY,
 		applied_at BIGINT NOT NULL DEFAULT FLOOR(EXTRACT(EPOCH FROM now()))::bigint

internal/db/db_test.go

@@ -0,0 +1,125 @@
+package db_test
+
+import (
+	"database/sql"
+	"fmt"
+	"net/url"
+	"os"
+	"strings"
+	"sync"
+	"testing"
+
+	"git.ryuvia.com/niklas/terdut-server/internal/db"
+
+	_ "github.com/jackc/pgx/v5/stdlib"
+)
+
+// TERDUT_TEST_DSN must point at a database the test role may create schemas
+// in; see internal/api/testdb_test.go for the fuller rationale this mirrors.
+// An unset DSN fails rather than skips, deliberately.
+const testDSNEnv = "TERDUT_TEST_DSN"
+
+// TestMigrate_ConcurrentCallersDoNotRace reproduces two replicas starting at
+// once against a brand-new, unmigrated schema: both call db.Migrate at the
+// same time. Before migrationLockKey, the loser could crash on a
+// duplicate-key insert into schema_migrations, or contend with the winner's
+// uncommitted DDL; with the advisory lock, one blocks until the other
+// finishes and both return cleanly.
+func TestMigrate_ConcurrentCallersDoNotRace(t *testing.T) {
+	dsn := os.Getenv(testDSNEnv)
+	if dsn == "" {
+		t.Fatalf("%s is not set: these tests need Postgres.\n"+
+			"Run `make test-db` for a local one, then\n"+
+			"  export %s=postgres://terdut:terdut@localhost:5432/terdut_test?sslmode=disable",
+			testDSNEnv, testDSNEnv)
+	}
+
+	schema := fmt.Sprintf("migrate_race_%d", os.Getpid())
+	admin, err := sql.Open("pgx", dsn)
+	if err != nil {
+		t.Fatalf("connect to %s: %v", testDSNEnv, err)
+	}
+	defer admin.Close()
+	if _, err := admin.Exec("CREATE SCHEMA " + schema); err != nil {
+		t.Fatalf("create schema %s: %v", schema, err)
+	}
+	t.Cleanup(func() {
+		cleanup, err := sql.Open("pgx", dsn)
+		if err != nil {
+			return
+		}
+		defer cleanup.Close()
+		if _, err := cleanup.Exec("DROP SCHEMA " + schema + " CASCADE"); err != nil {
+			t.Logf("drop schema %s: %v", schema, err)
+		}
+	})
+
+	scoped := withSearchPath(dsn, schema)
+
+	const callers = 2
+	errs := make([]error, callers)
+	var wg sync.WaitGroup
+	for i := range callers {
+		wg.Add(1)
+		go func(i int) {
+			defer wg.Done()
+			database, err := db.Open(scoped)
+			if err != nil {
+				errs[i] = fmt.Errorf("open: %w", err)
+				return
+			}
+			defer database.Close()
+			errs[i] = db.Migrate(database)
+		}(i)
+	}
+	wg.Wait()
+
+	for i, err := range errs {
+		if err != nil {
+			t.Fatalf("Migrate #%d: %v", i, err)
+		}
+	}
+
+	entries, err := os.ReadDir("migrations")
+	if err != nil {
+		t.Fatalf("read migrations dir: %v", err)
+	}
+	var want int
+	for _, e := range entries {
+		if !e.IsDir() && strings.HasSuffix(e.Name(), ".sql") {
+			want++
+		}
+	}
+
+	check, err := sql.Open("pgx", scoped)
+	if err != nil {
+		t.Fatalf("connect for verification: %v", err)
+	}
+	defer check.Close()
+
+	var got int
+	if err := check.QueryRow("SELECT COUNT(*) FROM schema_migrations").Scan(&got); err != nil {
+		t.Fatalf("count schema_migrations: %v", err)
+	}
+	if got != want {
+		t.Fatalf("schema_migrations has %d row(s) after two concurrent Migrate calls, want %d (one per migration file, no duplicates)", got, want)
+	}
+}
+
+// withSearchPath pins a DSN to one schema. Copied from
+// internal/api/testdb_test.go rather than shared: that helper lives in the
+// api_test package, a separate compiled package this one cannot import.
+func withSearchPath(dsn, schema string) string {
+	opt := "-csearch_path=" + schema
+
+	if strings.HasPrefix(dsn, "postgres://") || strings.HasPrefix(dsn, "postgresql://") {
+		u, err := url.Parse(dsn)
+		if err == nil {
+			q := u.Query()
+			q.Set("options", opt)
+			u.RawQuery = q.Encode()
+			return u.String()
+		}
+	}
+	return dsn + " options='" + opt + "'"
+}

internal/db/migrations/015_incident_service_account_actors.sql

@@ -0,0 +1,39 @@
+-- Service-account actors on incident mutations (terdut-server#25). A
+-- team-scoped service account acknowledging/resolving/snoozing/noting an
+-- incident is not a users row, so it cannot be written into
+-- acknowledged_by/incident_events.user_id — doing so either violates the
+-- users(id) FK (new rows) or, for incident_events.user_id, silently matches
+-- zero rows on delete. These columns are the service-account-shaped parallel
+-- to the existing human ones: nullable, mutually exclusive with their human
+-- counterpart, ON DELETE SET NULL so a deleted service account doesn't take
+-- the incident history with it.
+ALTER TABLE incidents
+    ADD COLUMN acknowledged_by_service_account_id BIGINT
+        REFERENCES service_accounts(id) ON DELETE SET NULL;
+
+ALTER TABLE incident_events
+    ADD COLUMN service_account_id BIGINT
+        REFERENCES service_accounts(id) ON DELETE SET NULL;
+
+-- At most one actor kind per row: both NULL ("the server acted") is valid,
+-- exactly one set is valid, both set is a bug this constraint refuses to
+-- store rather than silently accepting.
+ALTER TABLE incidents
+    ADD CONSTRAINT incidents_ack_actor_xor_chk CHECK (
+        acknowledged_by IS NULL OR acknowledged_by_service_account_id IS NULL
+    );
+
+ALTER TABLE incident_events
+    ADD CONSTRAINT incident_events_actor_xor_chk CHECK (
+        user_id IS NULL OR service_account_id IS NULL
+    );
+
+CREATE INDEX incidents_acknowledged_by_service_account_id_idx
+    ON incidents(acknowledged_by_service_account_id);
+CREATE INDEX incident_events_service_account_id_idx
+    ON incident_events(service_account_id);
+
+-- assigned_to_service_account_id is deliberately not added here: it would sit
+-- unpopulated until handleIncidentAssign itself tracks an actor, which is a
+-- separate, pre-existing gap (it records the assignee today, never the
+-- actor, for humans either) tracked in its own follow-up issue.

internal/db/migrations/016_rate_limit_counters.sql

@@ -0,0 +1,16 @@
+-- Backs the rate limiters (failed logins, sign-ups, OIDC/device start) with
+-- Postgres instead of an in-memory map, now that the server runs more than
+-- one replica in production (v0.37.0): a counter that only ever sees its own
+-- pod's traffic quietly let every one of these limits through multiplied by
+-- the replica count.
+--
+-- window_start is the start of the current fixed window for key, in the same
+-- "unix seconds" shape every other timestamp in this schema uses. The window
+-- resets rather than slides, matching the in-memory limiter it replaces:
+-- once a key's window is older than the limiter's window length, the next
+-- failure starts a fresh one instead of extending the stale one.
+CREATE TABLE rate_limit_counters (
+    key          TEXT   PRIMARY KEY,
+    window_start BIGINT NOT NULL,
+    count        INT    NOT NULL
+);

internal/db/migrations/017_api_key_expiry.sql

@@ -0,0 +1,7 @@
+-- Optional expiry on a user's own API keys. NULL (the existing default for
+-- every row already in this table) means "never expires" -- the same
+-- behavior these keys have always had, so no existing integration breaks.
+-- Service account keys are deliberately NOT touched: they are a different
+-- table, managed by automation, and already distinguished by their own
+-- "tdsa_" prefix.
+ALTER TABLE api_keys ADD COLUMN expires_at BIGINT;

internal/db/migrations/018_incident_event_actor.sql

@@ -0,0 +1,21 @@
+-- Who performed an assignment (terdut-server#35). On an 'assigned' event
+-- incident_events.user_id is the assignee, so the actor needs columns of its
+-- own. Only populated for 'assigned' events; every other event type keeps
+-- using user_id/service_account_id for the actor. Older 'assigned' rows stay
+-- NULL (the actor was never recorded). Same shape as migration 015: nullable,
+-- mutually exclusive, ON DELETE SET NULL.
+--
+-- assigned_to_service_account_id is still deliberately not added: making
+-- service accounts assignable is a separate change (request body, assignee
+-- picker, notifier, filters).
+ALTER TABLE incident_events
+    ADD COLUMN actor_user_id BIGINT REFERENCES users(id) ON DELETE SET NULL,
+    ADD COLUMN actor_service_account_id BIGINT REFERENCES service_accounts(id) ON DELETE SET NULL;
+
+ALTER TABLE incident_events
+    ADD CONSTRAINT incident_events_assign_actor_xor_chk CHECK (
+        actor_user_id IS NULL OR actor_service_account_id IS NULL
+    );
+
+CREATE INDEX incident_events_actor_user_id_idx ON incident_events(actor_user_id);
+CREATE INDEX incident_events_actor_service_account_id_idx ON incident_events(actor_service_account_id);

internal/models/incident.go

@@ -43,6 +43,13 @@ type Incident struct {
 	AcknowledgedByUser *string    `json:"acknowledged_by,omitempty"`
 	AcknowledgedAt     *time.Time `json:"acknowledged_at,omitempty"`
 
+	// AcknowledgedByServiceAccountID/Name are the service-account-shaped
+	// parallel to AcknowledgedByID/User above: mutually exclusive with it,
+	// populated when a service account (not a human) acknowledged this
+	// incident. See migration 015 and terdut-server#25.
+	AcknowledgedByServiceAccountID   *int64  `json:"acknowledged_by_service_account_id,omitempty"`
+	AcknowledgedByServiceAccountName *string `json:"acknowledged_by_service_account,omitempty"`
+
 	AssignedToID   *int64  `json:"assigned_to_id,omitempty"`
 	AssignedToUser *string `json:"assigned_to,omitempty"`
 
@@ -67,17 +74,36 @@ type Incident struct {
 // and is the only history this server keeps — alert rows are mutated in place.
 //
 // Type is one of: triggered, alert_added, alert_resolved, acknowledged,
-// unacknowledged, assigned, snoozed, unsnoozed, resolved, note. A nil UserID
-// means the server acted rather than a person.
+// unacknowledged, assigned, archived, unarchived, snoozed, unsnoozed,
+// resolved, note. UserID and
+// ServiceAccountID are mutually exclusive; both nil means the server acted
+// rather than any caller.
 type IncidentEvent struct {
-	ID         int64     `json:"id"`
-	IncidentID int64     `json:"incident_id"`
-	Type       string    `json:"type"`
-	UserID     *int64    `json:"user_id,omitempty"`
-	Username   *string   `json:"username,omitempty"`
-	AlertID    *int64    `json:"alert_id,omitempty"`
-	Detail     *string   `json:"detail,omitempty"`
-	CreatedAt  time.Time `json:"created_at"`
+	ID         int64   `json:"id"`
+	IncidentID int64   `json:"incident_id"`
+	Type       string  `json:"type"`
+	UserID     *int64  `json:"user_id,omitempty"`
+	Username   *string `json:"username,omitempty"`
+
+	// ServiceAccountID/Name are the service-account-shaped parallel to
+	// UserID/Username above: mutually exclusive with it, populated when a
+	// service account (not a human, and not nil-meaning-the-server-acted)
+	// performed this event. Named Name, not Username — a ServiceAccount has
+	// a Name field, not a Username. See migration 015 and terdut-server#25.
+	ServiceAccountID   *int64  `json:"service_account_id,omitempty"`
+	ServiceAccountName *string `json:"service_account_name,omitempty"`
+
+	// Actor* name who performed an 'assigned' event, whose UserID is the
+	// assignee. Mutually exclusive; unset on every other event type and on
+	// assignments made before migration 018. See terdut-server#35.
+	ActorUserID             *int64  `json:"actor_user_id,omitempty"`
+	ActorUsername           *string `json:"actor_username,omitempty"`
+	ActorServiceAccountID   *int64  `json:"actor_service_account_id,omitempty"`
+	ActorServiceAccountName *string `json:"actor_service_account_name,omitempty"`
+
+	AlertID   *int64    `json:"alert_id,omitempty"`
+	Detail    *string   `json:"detail,omitempty"`
+	CreatedAt time.Time `json:"created_at"`
 }
 
 // SimilarIncident is an earlier, resolved incident with the same signature as

internal/models/user.go

@@ -37,5 +37,13 @@ type APIKey struct {
 	Name       string     `json:"name"`
 	CreatedAt  time.Time  `json:"created_at"`
 	LastUsedAt *time.Time `json:"last_used_at,omitempty"`
-	Key        string     `json:"key,omitempty"` // populated only on creation, never stored
+
+	// ExpiresAt is nil for a key that never expires, which is every key
+	// created before this field existed and still the default for a new one
+	// unless its creator asks otherwise (see handleCreateAPIKey's
+	// expires_in_days). AuthMiddleware stops accepting a key once this
+	// passes; nothing deletes the row for it.
+	ExpiresAt *time.Time `json:"expires_at,omitempty"`
+
+	Key string `json:"key,omitempty"` // populated only on creation, never stored
 }

internal/web/static/app.css

@@ -10,11 +10,11 @@
   --surface: #ffffff;
   --surface-2: #eff1f4;
   --surface-hover: #f7f8fa;
-  --border: #e2e5ea;
-  --border-strong: #cfd3da;
+  --border: #d9dde4;
+  --border-strong: #c5cad3;
   --text: #16181d;
   --muted: #5b626e;
-  --faint: #8a909b;
+  --faint: #676d79;
 
   --accent: #2f5bd3;
   --accent-text: #ffffff;
@@ -72,17 +72,21 @@
   --safe-bottom: env(safe-area-inset-bottom, 0px);
 }
 
+/* Dark palette. Applied by the OS preference unless the user chose Light in
+   Account (data-theme="light" on <html>, set by js/theme.js), and always when
+   they chose Dark. The two blocks must stay identical: CSS has no way to share
+   a declaration list between a media query and an attribute selector. */
 @media (prefers-color-scheme: dark) {
-  :root {
+  :root:not([data-theme="light"]) {
     --bg: #0f1115;
-    --surface: #171a20;
-    --surface-2: #1f232b;
-    --surface-hover: #1c2027;
-    --border: #2a2f38;
-    --border-strong: #394050;
+    --surface: #1a1e26;
+    --surface-2: #232834;
+    --surface-hover: #20252e;
+    --border: #343b49;
+    --border-strong: #444d5f;
     --text: #e7e9ed;
     --muted: #a0a7b3;
-    --faint: #737a87;
+    --faint: #8a92a0;
 
     --accent: #6d8ff0;
     --accent-text: #0b0d12;
@@ -108,6 +112,42 @@
     --shadow-lg: 0 16px 40px rgb(0 0 0 / 55%);
   }
 }
+:root[data-theme="dark"] {
+  --bg: #0f1115;
+  --surface: #1a1e26;
+  --surface-2: #232834;
+  --surface-hover: #20252e;
+  --border: #343b49;
+  --border-strong: #444d5f;
+  --text: #e7e9ed;
+  --muted: #a0a7b3;
+  --faint: #8a92a0;
+
+  --accent: #6d8ff0;
+  --accent-text: #0b0d12;
+  --accent-soft: #1d2640;
+
+  --crit: #ff6b61;
+  --crit-soft: #3a1c1b;
+  --warn: #f0b140;
+  --warn-soft: #362a14;
+  --info: #74a3ff;
+  --info-soft: #1a2640;
+  --ok: #4cc488;
+  --ok-soft: #15301f;
+  --snooze: #a89bff;
+  --snooze-soft: #262245;
+
+  --teal: #4fc2d4;
+  --teal-soft: #0f2e33;
+  --pink: #f07fb8;
+  --pink-soft: #3a1c2d;
+
+  --shadow: 0 1px 2px rgb(0 0 0 / 40%);
+  --shadow-lg: 0 16px 40px rgb(0 0 0 / 55%);
+}
+:root[data-theme="light"] { color-scheme: light; }
+:root[data-theme="dark"] { color-scheme: dark; }
 
 *, *::before, *::after { box-sizing: border-box; }
 [hidden] { display: none !important; }
@@ -237,6 +277,10 @@ input:focus, textarea:focus { outline: none; border-color: var(--accent); box-sh
 .open-pill.has-triggered { background: var(--crit-soft); color: var(--crit); }
 .open-pill.has-triggered::before { background: var(--crit); }
 .open-pill.all-acked::before { background: var(--warn); }
+/* "All clear" is a status, not an action: green with a check, no dot. */
+.open-pill.all-clear { background: var(--ok-soft); color: var(--ok); }
+.open-pill.all-clear::before { content: none; }
+.open-pill .icon { width: 14px; height: 14px; }
 
 /* Bottom tab bar on phones (Queue, On-call, Alerts, Team, More); becomes the
    left sidebar from 900px, where the desktop block below redeclares display
@@ -290,13 +334,13 @@ input:focus, textarea:focus { outline: none; border-color: var(--accent); box-sh
 
 .nav-team-selector,
 .team-selector-mobile {
-  display: inline-flex; align-items: center; gap: 8px;
+  align-items: center; gap: 8px;
   border: 1px solid var(--border-strong); border-radius: 999px;
   background: var(--surface); color: var(--text);
   font-size: 13px; font-weight: 600; cursor: pointer;
   padding: 4px 12px; max-width: 100%;
 }
-.team-selector-mobile { padding: 4px 10px; font-size: 12px; max-width: 120px; }
+.team-selector-mobile { display: inline-flex; padding: 4px 10px; font-size: 12px; max-width: 120px; }
 .team-selector-label { overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
 .team-selector-chevron { width: 14px; height: 14px; flex: none; color: var(--faint); margin-left: -2px; }
 
@@ -340,7 +384,8 @@ input:focus, textarea:focus { outline: none; border-color: var(--accent); box-sh
   background: var(--surface); color: var(--muted);
   font-size: 13px; font-weight: 600; cursor: pointer;
 }
-.chip[aria-selected="true"] { background: var(--text); border-color: var(--text); color: var(--bg); }
+.chip[aria-selected="true"], .chip[aria-checked="true"] { background: var(--text); border-color: var(--text); color: var(--bg); }
+.theme-picker { padding: 0 0 8px; }
 .chip .count { margin-left: 4px; opacity: 0.7; }
 /* An overlay, not a flex item: absolute against .chips' own (non-scrolling)
    box stays flush with its real right edge regardless of scroll position,
@@ -583,6 +628,10 @@ details[open] > summary { margin-bottom: 8px; }
 }
 .actionbar .btn { min-height: 48px; }
 .actionbar .btn-primary { flex: 1; font-size: 16px; }
+/* The extra actions that only phones hide behind "More" — see actionBar() and
+   extraActions() in incident.js. Hidden by default; the desktop block below
+   shows them and hides the now-redundant More button instead. */
+.action-extra { display: none; }
 
 .detail-placeholder {
   display: grid; place-items: center; height: 100%;
@@ -646,7 +695,6 @@ details[open] > summary { margin-bottom: 8px; }
 .page-head { display: flex; align-items: center; justify-content: space-between; gap: 8px; margin: 16px auto 12px; }
 .page-head h2 { font-size: 13px; font-weight: 700; text-transform: uppercase; letter-spacing: 0.06em; color: var(--muted); }
 
-.now-card { display: flex; align-items: center; gap: 14px; padding: 16px; margin-top: 16px; }
 .avatar {
   flex: none; display: grid; place-items: center;
   width: 44px; height: 44px; border-radius: 50%;
@@ -654,8 +702,6 @@ details[open] > summary { margin-bottom: 8px; }
   font-weight: 750; font-size: 17px; text-transform: uppercase;
 }
 .avatar.none { background: var(--surface-2); color: var(--faint); }
-.now-label { color: var(--muted); font-size: 13px; font-weight: 600; }
-.now-name { font-size: 20px; font-weight: 750; }
 .you { color: var(--accent); font-weight: 650; font-size: 13px; margin-left: 6px; }
 /* Oncall's own "you" indicator only (see you() in oncall.js) — a pill badge
    is easier to spot there than this plain accent-coloured text. */
@@ -669,25 +715,49 @@ details[open] > summary { margin-bottom: 8px; }
    caught by the unrelated .label > span styling meant for label chips. */
 .week-nav .week-label { font-size: 14px; font-weight: 650; min-width: 11.5em; text-align: center; white-space: nowrap; }
 .week-label small { color: var(--faint); font-weight: 600; font-size: 11px; margin-left: 2px; }
-.days { list-style: none; margin: 0; padding: 0; }
-.day { display: grid; grid-template-columns: 3.2em 4.2em 1fr; align-items: center; gap: 8px; min-height: 50px; padding: 0 14px; }
-.day + .day { border-top: 1px solid var(--border); }
-.day-name { font-weight: 650; }
-.day-date { color: var(--faint); font-size: 13px; }
-.day-who { overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
-.day-who.nobody { color: var(--faint); font-style: italic; }
-/* A run of several days held by the same person (or left empty), replacing
-   what used to be one identical row per day — see weekRuns() in oncall.js. */
-.day.range { grid-template-columns: 1fr auto; }
-.day-range { font-weight: 650; }
-.day.today { background: var(--accent-soft); }
-.day.today:first-child { border-radius: var(--radius) var(--radius) 0 0; }
-.day.today:last-child { border-radius: 0 0 var(--radius) var(--radius); }
-.day.today .day-name, .day.today .day-range { color: var(--accent); }
-.day.past { opacity: 0.6; }
-/* Highlights whichever row is yours, same soft tint as .today — they already
-   read fine layered (today's own row is almost always one of yours too). */
-.day.mine { background: var(--accent-soft); }
+.oncall-title h1 { font-size: 24px; font-weight: 600; letter-spacing: -0.01em; margin: 8px 0 0; }
+.oncall-title p { margin: 2px 0 0; font-size: 14px; }
+.oncall-grid { display: grid; gap: 0; }
+.oncall-main, .oncall-side { min-width: 0; }
+.hero-list { display: grid; gap: 12px; }
+
+.hero { display: grid; gap: 14px; padding: 18px; margin-top: 16px; }
+.hero-top { display: flex; align-items: center; gap: 16px; }
+.hero-label { font-size: 12px; font-weight: 700; text-transform: uppercase; letter-spacing: 0.07em; color: var(--muted); }
+.hero-avatar { width: 64px; height: 64px; font-size: 24px; }
+.hero-name { font-size: 24px; font-weight: 650; letter-spacing: -0.01em; display: flex; align-items: center; flex-wrap: wrap; gap: 4px; }
+.hero-until { color: var(--muted); font-size: 15px; margin-top: 2px; }
+
+.week-card { margin-top: 16px; padding: 16px 12px 12px; }
+.week-head { display: flex; align-items: center; justify-content: space-between; gap: 8px; margin-bottom: 12px; padding: 0 4px; }
+.week-head h2 { margin: 0; font-size: 15px; font-weight: 650; }
+/* Bordered, so the arrows read as buttons. 40px: a touch target without
+   crowding the date between them. */
+.week-arrow { width: 40px; min-height: 40px; border: 1px solid var(--border-strong); }
+.strip { list-style: none; margin: 0; padding: 0; display: grid; grid-template-columns: repeat(7, minmax(0, 1fr)); gap: 2px; }
+.strip-day {
+  position: relative; min-width: 0;
+  display: flex; flex-direction: column; align-items: center; gap: 4px;
+  padding: 10px 0 8px; border: 1px solid transparent; border-radius: var(--radius);
+}
+.strip-name { font-size: 12px; font-weight: 650; text-transform: uppercase; letter-spacing: 0.04em; color: var(--muted); }
+.strip-date { font-size: 15px; font-weight: 650; }
+.strip-avatar { width: 32px; height: 32px; font-size: 13px; }
+.strip-who { max-width: 100%; padding: 0 2px; font-size: 12px; color: var(--muted); overflow: hidden; text-overflow: ellipsis; white-space: nowrap; }
+.strip-who.nobody { font-style: italic; color: var(--faint); }
+.strip-day.past { opacity: 0.6; }
+/* "Current", not "selected": a neutral tint, a bar and aria-current, so it
+   reads as a marker and not as a pressed button. The accent stays for things
+   you can act on. */
+.strip-day.today { background: var(--surface-2); border-color: var(--border-strong); }
+.strip-day.today::before {
+  content: ""; position: absolute; top: -1px; left: 12px; right: 12px; height: 3px;
+  border-radius: 0 0 3px 3px; background: var(--text);
+}
+.strip-day.today .strip-name { color: var(--text); }
+.strip-legend { display: flex; flex-wrap: wrap; gap: 4px 14px; margin-top: 10px; padding: 10px 4px 0; border-top: 1px solid var(--border); font-size: 13px; color: var(--muted); }
+.strip-legend span { display: inline-flex; align-items: center; gap: 6px; }
+.strip-legend .strip-dot { width: 8px; height: 8px; border-radius: 50%; background: currentColor; }
 
 .shift-list { list-style: none; margin: 0; padding: 0; }
 .shift-list li { display: flex; justify-content: space-between; padding: 12px 14px; }
@@ -773,7 +843,15 @@ kbd {
   .pane-list .chips { flex-wrap: wrap; overflow-x: visible; }
   .pane-list .chip-sep { display: none; }
   .chips-fade { display: none; }
-  .view-queue:not(.has-detail) .pane-detail { display: block; }
+  /* With nothing selected there is no detail to show next to, so the list
+     takes the whole row instead of leaving the second column as dead space
+     around the placeholder text. Selecting an incident (.has-detail) drops
+     back to the base minmax(340,420) 1fr rule above. */
+  .view-queue:not(.has-detail) { grid-template-columns: 1fr; }
+  .view-queue:not(.has-detail) .pane-list { border-right: 0; }
+  /* Full width reads better capped than edge-to-edge on a very wide monitor,
+     matching .detail's own cap below. */
+  .view-queue:not(.has-detail) .list { max-width: 900px; margin: 0 auto; }
 
   /* On desktop the list stays visible next to the detail. */
   .app.detail-open .nav { display: flex; }
@@ -787,8 +865,12 @@ kbd {
   .actionbar {
     position: sticky; bottom: 0;
     padding: 12px 32px;
+    flex-wrap: wrap;
   }
   .actionbar .btn-primary { flex: 0 1 240px; }
+  /* Room enough to show every action, so More is fully redundant here. */
+  .action-extra { display: inline-flex; }
+  .more-btn { display: none; }
 
   .sheet {
     width: min(440px, calc(100% - 32px));
@@ -829,6 +911,11 @@ kbd {
 .status-table th, .status-table td { white-space: nowrap; }
 .status-table td.wrap { white-space: normal; min-width: 12em; }
 .status-table .source-row td { border-bottom-style: dashed; }
+.status-table tr.clickable { cursor: pointer; }
+.status-table tr.clickable:hover td, .status-table tr.clickable:focus-visible td { background: rgba(127, 127, 127, .1); }
+.switch-facts { display: grid; grid-template-columns: max-content 1fr; gap: 6px 16px; margin: 12px 0; }
+.switch-facts dt { opacity: .7; }
+.switch-facts dd { margin: 0; }
 .status-table .source-row td:first-child { padding-left: 16px; }
 .source-labels { display: flex; flex-wrap: wrap; gap: 4px; align-items: center; }
 
@@ -1017,6 +1104,11 @@ button.rota-week:hover { background: var(--surface-2); color: var(--text); }
   overflow-x: auto; scrollbar-width: none;
 }
 .subnav::-webkit-scrollbar { display: none; }
+/* A fade on the right edge says "there is more" where the strip overflows;
+   the desktop width fits every entry, so it is phone-only. */
+@media (max-width: 899px) {
+  .subnav { -webkit-mask-image: linear-gradient(90deg, #000 calc(100% - 28px), transparent); mask-image: linear-gradient(90deg, #000 calc(100% - 28px), transparent); }
+}
 .subnav-link {
   flex: none;
   padding: 8px 12px; margin-bottom: -1px;
@@ -1076,3 +1168,10 @@ button.rota-week:hover { background: var(--surface-2); color: var(--text); }
 @media (min-width: 900px) {
   .stat-tiles { grid-template-columns: repeat(3, 1fr); }
 }
+
+/* On-call, wide: the hero and the week on the left, your shifts on the right.
+   .view-page > * caps its children at 760px, so the grid lifts that. */
+@media (min-width: 900px) {
+  #view-oncall > .oncall-grid { max-width: 1080px; grid-template-columns: minmax(0, 1.6fr) minmax(0, 1fr); gap: 0 24px; align-items: start; }
+  #view-oncall > .oncall-title { max-width: 1080px; }
+}

internal/web/static/index.html

@@ -14,6 +14,7 @@
   <link rel="icon" href="/icon.svg" type="image/svg+xml">
   <link rel="apple-touch-icon" href="/apple-touch-icon.png">
   <link rel="stylesheet" href="/app.css">
+  <script src="/js/theme.js"></script>
   <script type="module" src="/js/app.js"></script>
 </head>
 <body>

internal/web/static/js/account.js

@@ -28,15 +28,37 @@ function render() {
 
     ...passwordSection(user, hasPassword),
 
+    h('div', { class: 'page-head' }, h('h2', { text: 'Appearance' })),
+    themePicker(),
+
     h('div', { class: 'only-desktop' },
       h('div', { class: 'page-head' }, h('h2', { text: 'Keyboard' })),
       h('div', { class: 'card' }, shortcuts())),
 
     h('div', { class: 'page-head' }),
-    h('button', { class: 'btn btn-block', type: 'button', onclick: signOut }, icon('logout'), 'Sign out'),
+    // Wrapped in a div: .btn is inline-flex, and only a block-level element
+    // picks up .view-page > *'s margin:auto centering (see app.css:316).
+    h('div', {}, h('button', { class: 'btn btn-block', type: 'button', onclick: signOut }, icon('logout'), 'Sign out')),
   );
 }
 
+// System / Light / Dark. Per browser, not per account: it lives in
+// localStorage (see js/theme.js), the same place the selected team does.
+const THEMES = [['system', 'System'], ['light', 'Light'], ['dark', 'Dark']];
+
+function themePicker() {
+  const theme = window.terdutTheme;
+  const buttons = THEMES.map(([value, text]) => h('button', {
+    class: 'chip', type: 'button', role: 'radio', text,
+    onclick: () => { theme.set(value); sync(); },
+  }));
+  const sync = () => buttons.forEach((b, i) => b.setAttribute('aria-checked', String(THEMES[i][0] === theme.get())));
+  sync();
+  return h('div', { class: 'card card-pad' },
+    h('div', { class: 'chips theme-picker', role: 'radiogroup', 'aria-label': 'Theme' }, ...buttons),
+    h('p', { class: 'muted small', text: 'System follows your device. This applies to this browser only.' }));
+}
+
 // Where this user's pages go. The onboarding checklist's first step sends
 // people here for it, and until now there was nothing here to send them to:
 // the topic could only be set with curl or by an administrator.

internal/web/static/js/admin.js

@@ -102,13 +102,18 @@ function section() {
 // buttons, because these are four URLs: app.js intercepts the click, the
 // browser's Back walks them, and a reload lands where you were.
 function subnav() {
-  return h('nav', { class: 'subnav', 'aria-label': 'Administration' },
+  const nav = h('nav', { class: 'subnav', 'aria-label': 'Administration' },
     TABS.map((t) => h('a', {
       class: 'subnav-link',
       href: t.path,
       text: t.label,
       'aria-current': t.tab === tab ? 'page' : null,
     })));
+  // On a phone the strip overflows; bring the open section into view so a
+  // tab past the edge (Sources, Switches) is never the one that is hidden.
+  requestAnimationFrame(() => nav.querySelector('[aria-current]')
+    ?.scrollIntoView({ inline: 'center', block: 'nearest' }));
+  return nav;
 }
 
 // --- overview --------------------------------------------------------------

internal/web/static/js/adminteam.js

@@ -96,7 +96,9 @@ function render() {
 }
 
 function backLink() {
-  return h('a', { class: 'back-link', href: '/admin/teams' }, icon('chevronLeft'), h('span', { text: 'Teams' }));
+  // Wrapped in a div: .back-link is inline-flex, and only a block-level
+  // element picks up .view-page > *'s margin:auto centering (app.css:316).
+  return h('div', {}, h('a', { class: 'back-link', href: '/admin/teams' }, icon('chevronLeft'), h('span', { text: 'Teams' })));
 }
 
 // --- identity --------------------------------------------------------------

internal/web/static/js/adminuser.js

@@ -83,7 +83,9 @@ function render() {
 }
 
 function backLink() {
-  return h('a', { class: 'back-link', href: '/admin/users' }, icon('chevronLeft'), h('span', { text: 'Users' }));
+  // Wrapped in a div: .back-link is inline-flex, and only a block-level
+  // element picks up .view-page > *'s margin:auto centering (app.css:316).
+  return h('div', {}, h('a', { class: 'back-link', href: '/admin/users' }, icon('chevronLeft'), h('span', { text: 'Users' })));
 }
 
 // --- identity --------------------------------------------------------------

internal/web/static/js/api.js

@@ -150,6 +150,8 @@ export const deleteIntegration = (id, integrationID) =>
 export const deadmanSwitches = (id) => call('GET', `/teams/${id}/deadman/switches`);
 export const createDeadmanSwitch = (id, body) =>
   call('POST', `/teams/${id}/deadman/switches`, { body });
+export const updateDeadmanSwitch = (id, switchID, body) =>
+  call('PUT', `/teams/${id}/deadman/switches/${switchID}`, { body });
 export const deleteDeadmanSwitch = (id, switchID) =>
   call('DELETE', `/teams/${id}/deadman/switches/${switchID}`);
 

internal/web/static/js/app.js

@@ -202,7 +202,8 @@ function updateBadges() {
 
   const pill = $('open-pill');
   pill.hidden = false;
-  pill.textContent = open ? `${open} open` : 'All clear';
+  pill.replaceChildren(...(open ? [`${open} open`] : [ui.icon('checkCircle'), 'All clear']));
+  pill.classList.toggle('all-clear', open === 0);
   pill.classList.toggle('has-triggered', triggered > 0);
   pill.classList.toggle('all-acked', open > 0 && triggered === 0);
 

internal/web/static/js/incident.js

@@ -123,6 +123,16 @@ function who(id, name) {
   return name || 'someone';
 }
 
+// ackActorLabel renders whoever acknowledged inc, human or service account —
+// the two are mutually exclusive (migration 015), and a service account is a
+// credential, not "you" or "nobody", so it gets its own branch rather than
+// going through who()'s id-vs-myID() check.
+function ackActorLabel() {
+  if (inc.acknowledged_by_id != null) return who(inc.acknowledged_by_id, inc.acknowledged_by);
+  if (inc.acknowledged_by_service_account_id != null) return inc.acknowledged_by_service_account || 'a service account';
+  return null;
+}
+
 function facts() {
   const rows = [];
   const add = (k, ...v) => rows.push(h('dt', { text: k }), h('dd', {}, ...v));
@@ -132,8 +142,7 @@ function facts() {
   // take reading top to bottom to piece together.
   const elapsedTo = inc.resolved_at ? Date.parse(inc.resolved_at) : Date.now();
   const responsible = inc.assigned_to_id != null ? who(inc.assigned_to_id, inc.assigned_to)
-    : inc.acknowledged_by_id != null ? who(inc.acknowledged_by_id, inc.acknowledged_by)
-      : 'Unassigned';
+    : ackActorLabel() || 'Unassigned';
   rows.push(h('dt', { text: 'At a glance' }), h('dd', { class: 'fact-summary' },
     h('span', { class: 'fact-chip' }, icon('clock', 'icon fact-icon'), duration(elapsedTo - Date.parse(inc.triggered_at))),
     inc.severity && badge(inc.severity, `plain ${severityClass(inc.severity)}`),
@@ -142,7 +151,7 @@ function facts() {
 
   add('Triggered', when(inc.triggered_at), h('span', { class: 'sub', text: ` · ${ago(inc.triggered_at)}` }));
   if (inc.acknowledged_at) {
-    add('Acknowledged', `${who(inc.acknowledged_by_id, inc.acknowledged_by)} · ${when(inc.acknowledged_at)}`);
+    add('Acknowledged', `${ackActorLabel()} · ${when(inc.acknowledged_at)}`);
   }
   add('Assigned', inc.assigned_to_id != null ? who(inc.assigned_to_id, inc.assigned_to) : 'Unassigned');
   if (inc.status !== 'resolved' && isFuture(inc.snoozed_until)) {
@@ -206,9 +215,28 @@ function alertItem(a) {
 
 // ---------- timeline ----------
 
+// actorLabel renders whoever performed ev, human or service account — the
+// two are mutually exclusive (migration 015). null means the server acted:
+// ev.user_id == null no longer means that by itself, now that a service
+// account's events also leave it null.
+function actorLabel(ev, named = false) {
+  if (ev.user_id != null) return named ? (ev.username || 'someone') : who(ev.user_id, ev.username);
+  if (ev.service_account_id != null) return ev.service_account_name || 'a service account';
+  return null;
+}
+
+// assignerLabel is actorLabel for an 'assigned' event, whose user_id is the
+// assignee: the person who made the assignment is in the actor_* fields
+// (null for assignments from before they were recorded).
+function assignerLabel(ev, named = false) {
+  if (ev.actor_user_id != null) return named ? (ev.actor_username || 'someone') : who(ev.actor_user_id, ev.actor_username);
+  if (ev.actor_service_account_id != null) return ev.actor_service_account_name || 'a service account';
+  return null;
+}
+
 // named spells users out instead of "you", for text that leaves this page.
 function eventText(ev, named = false) {
-  const person = ev.user_id != null ? (named ? ev.username || 'someone' : who(ev.user_id, ev.username)) : null;
+  const person = actorLabel(ev, named);
   const strong = (t) => h('span', { class: 'who', text: t || 'someone' });
   const alertName = () => {
     const a = (inc.alerts || []).find((x) => x.id === ev.alert_id);
@@ -220,7 +248,12 @@ function eventText(ev, named = false) {
     case 'alert_resolved': return [`Alert resolved: ${alertName()}`];
     case 'acknowledged': return [strong(person), ' acknowledged'];
     case 'unacknowledged': return [strong(person), ' cleared the acknowledgement'];
-    case 'assigned': return ['Assigned to ', strong(person)];
+    case 'assigned': {
+      const by = assignerLabel(ev, named);
+      return by ? ['Assigned to ', strong(person), ' by ', strong(by)] : ['Assigned to ', strong(person)];
+    }
+    case 'archived': return [strong(person), ' archived the incident'];
+    case 'unarchived': return [strong(person), ' unarchived the incident'];
     case 'snoozed': return [strong(person), ` snoozed until ${ev.detail ? when(ev.detail) : '…'}`];
     case 'unsnoozed': return [strong(person), ' ended the snooze'];
     case 'resolved': return person ? [strong(person), ' resolved the incident'] : ['Resolved: every alert stopped firing'];
@@ -453,9 +486,25 @@ function quickActions() {
   return div;
 }
 
+// Desktop has room to show what a phone folds into the More sheet below — see
+// the .action-extra/.more-btn rules in app.css. Resolved/archived incidents
+// already say everything via primaryAction()/secondaryAction(), so there is
+// nothing extra to surface for them.
+function extraActions() {
+  if (!isOpen()) return [];
+  const out = [
+    h('button', { class: 'btn btn-sm action-extra', type: 'button', onclick: assign }, icon('user'), 'Assign…'),
+    h('button', { class: 'btn btn-sm action-extra', type: 'button', onclick: addNote }, icon('note'), 'Add note…'),
+  ];
+  out.push(inc.status === 'acknowledged'
+    ? h('button', { class: 'btn btn-sm action-extra', type: 'button', onclick: unacknowledge }, icon('undo'), 'Clear ack')
+    : h('button', { class: 'btn btn-sm action-extra', type: 'button', onclick: resolve }, icon('checkCircle'), 'Resolve…'));
+  return out;
+}
+
 function actionBar() {
-  const more = h('button', { class: 'btn', type: 'button', 'aria-label': 'More actions', onclick: moreMenu }, icon('more'), 'More');
-  const bar = h('div', { class: 'actionbar' }, primaryAction(), secondaryAction(), more);
+  const more = h('button', { class: 'btn more-btn', type: 'button', 'aria-label': 'More actions', onclick: moreMenu }, icon('more'), 'More');
+  const bar = h('div', { class: 'actionbar' }, primaryAction(), secondaryAction(), ...extraActions(), more);
   if (busy) for (const b of bar.querySelectorAll('button')) b.disabled = true;
   return bar;
 }

internal/web/static/js/oncall.js

@@ -1,5 +1,6 @@
 // On-call: who is on duty now, the week around it, and your own next shifts.
-// Read-only for now; the TUI edits the schedule.
+// Read-only for now; the TUI edits the schedule, so there is no add or swap
+// button here.
 //
 // One team's rota at a time — the viewer's first team, since a viewer in one
 // team has nothing to choose between. "On call now" is the exception and shows
@@ -59,11 +60,17 @@ function render() {
     clear(view(), error ? h('div', { class: 'load-error', text: error }) : spinner());
     return;
   }
+  const team = currentTeam();
   clear(view(),
+    // The top bar carries the title on a phone; the desktop has none.
+    h('div', { class: 'only-desktop oncall-title' },
+      h('h1', { text: 'On-call' }),
+      team && h('p', { class: 'muted', text: `${team.name} · who is on call, the week ahead and your shifts` })),
     error && h('div', { class: 'load-error', text: `Showing older data: ${error}` }),
-    nowCard(),
-    weekCard(),
-    myShifts(),
+    h('div', { class: 'oncall-grid' },
+      h('div', { class: 'oncall-main' }, nowCard(), weekCard()),
+      h('div', { class: 'oncall-side' }, myShifts()),
+    ),
   );
 }
 
@@ -78,75 +85,73 @@ function nowCard() {
   const entries = data.now || [];
   const showTeam = entries.length > 1;
   if (entries.length === 0) {
-    return h('div', { class: 'card now-card' },
-      h('div', { class: 'avatar none', text: '–' }),
-      h('div', {},
-        h('div', { class: 'now-label', text: 'On call now' }),
-        h('div', { class: 'now-name', text: 'Nobody' }),
-      ),
-    );
+    return h('div', { class: 'card hero' },
+      h('div', { class: 'hero-label', text: 'On call now' }),
+      h('div', { class: 'hero-top' },
+        h('div', { class: 'avatar hero-avatar none', text: '–' }),
+        h('div', { class: 'hero-name', text: 'Nobody' })));
   }
-  return h('div', {}, ...entries.map((n) =>
-    h('div', { class: 'card now-card' },
-      h('div', { class: 'avatar', text: initial(n.username) }),
-      h('div', {},
-        h('div', {
-          class: 'now-label',
-          text: showTeam ? `On call now · ${n.team_name}` : 'On call now',
-        }),
-        h('div', { class: 'now-name' }, n.username, you(n.user_id)),
-      ),
-    )));
+  return h('div', { class: 'hero-list' }, ...entries.map((n) => {
+    const until = shiftEnd(n);
+    return h('div', { class: 'card hero' },
+      h('div', { class: 'hero-label', text: showTeam ? `On call now · ${n.team_name}` : 'On call now' }),
+      h('div', { class: 'hero-top' },
+        h('div', { class: 'avatar hero-avatar', text: initial(n.username) }),
+        h('div', {},
+          h('div', { class: 'hero-name' }, n.username, you(n.user_id)),
+          until && h('div', { class: 'hero-until', text: until }))));
+  }));
 }
 
-// Groups the week's 7 days into runs held by the same person (or the same
-// empty slot) — the week's own version of the consecutive-day grouping
-// myShifts does for a single person's own dates, below. Seven identical rows
-// for one person all week collapses to the one bar this way.
-function weekRuns(byDate) {
-  const runs = [];
-  for (let i = 0; i < 7; i++) {
-    const date = isoDate(addDays(weekStart, i));
-    const e = byDate.get(date) || null;
-    const uid = e ? e.user_id : null;
-    const last = runs[runs.length - 1];
-    if (last && last.uid === uid) last.to = date;
-    else runs.push({ uid, entry: e, from: date, to: date });
-  }
-  return runs;
+// "until Mon 12 Oct · ends in 3d 20h", for the current team only: the shift's
+// end is read off the rota (the run of consecutive days from today held by the
+// same person), and that is only loaded for one team. The same midnight
+// boundary the "Current shift" row below uses.
+function shiftEnd(n) {
+  const team = currentTeam();
+  if (!team || n.team_id !== team.id) return null;
+  const mine = new Map(data.upcoming.map((e) => [e.date, e.user_id]));
+  let day = new Date();
+  if (mine.get(isoDate(day)) !== n.user_id) return null;
+  while (mine.get(isoDate(addDays(day, 1))) === n.user_id) day = addDays(day, 1);
+  // Still holding the last day loaded: the shift may run on past it.
+  if (isoDate(day) >= data.upcoming.reduce((m, e) => (e.date > m ? e.date : m), '')) return null;
+  const end = addDays(parse(isoDate(day)), 1);
+  return `until ${dayName.format(end)} ${dayDate.format(end)} · ends in ${duration(end - Date.now())}`;
 }
 
+// A stable colour per person from the six-colour rcN palette. The week page
+// has no member list to take an index from (team.js does), so the id decides.
+const personClass = (userID) => `rc${(userID % 6) + 1}`;
+
+// The week as seven cells, Monday to Sunday, each showing who holds that day.
+// A hand-over in the middle of the week is visible without reading anything.
 function weekCard() {
   const byDate = new Map(data.week.map((e) => [e.date, e]));
   const today = isoDate(new Date());
   const mine = myID();
-  const days = weekRuns(byDate).map((r) => {
-    const single = r.from === r.to;
-    const cls = [
-      'day',
-      single ? '' : 'range',
-      r.from <= today && today <= r.to ? 'today' : '',
-      r.to < today ? 'past' : '',
-      r.uid === mine ? 'mine' : '',
-    ].filter(Boolean).join(' ');
-    const label = single
-      ? [h('span', { class: 'day-name', text: dayName.format(parse(r.from)) }),
-        h('span', { class: 'day-date', text: dayDate.format(parse(r.from)) })]
-      : [h('span', {
-        class: 'day-range',
-        text: `${dayName.format(parse(r.from))} ${dayDate.format(parse(r.from))} – ${dayName.format(parse(r.to))} ${dayDate.format(parse(r.to))}`,
-      })];
-    return h('li', { class: cls },
-      ...label,
-      h('span', { class: `day-who ${r.entry ? '' : 'nobody'}` }, r.entry ? r.entry.username : 'nobody', r.entry && you(r.entry.user_id)),
-    );
-  });
+  const seen = new Map();
+  const cells = [];
+  for (let i = 0; i < 7; i++) {
+    const d = addDays(weekStart, i);
+    const date = isoDate(d);
+    const e = byDate.get(date) || null;
+    if (e) seen.set(e.user_id, e.username);
+    cells.push(h('li', {
+      class: ['strip-day', date === today && 'today', date < today && 'past'].filter(Boolean).join(' '),
+      'aria-current': date === today ? 'date' : null,
+    },
+    h('span', { class: 'strip-name', text: dayName.format(d) }),
+    h('span', { class: 'strip-date', text: String(d.getDate()) }),
+    h('span', { class: `avatar strip-avatar ${e ? personClass(e.user_id) : 'none'}`, text: e ? initial(e.username) : '–' }),
+    h('span', { class: `strip-who${e ? '' : ' nobody'}`, text: e ? (e.user_id === mine ? 'You' : e.username) : 'nobody' })));
+  }
   const thisWeek = isoDate(weekStart) === isoDate(mondayOf(new Date()));
-  return [
-    h('div', { class: 'page-head' },
+  return h('div', { class: 'card week-card' },
+    h('div', { class: 'week-head' },
       h('h2', { text: thisWeek ? 'This week' : 'Week' }),
       h('div', { class: 'week-nav' },
-        h('button', { class: 'btn btn-ghost btn-icon', type: 'button', 'aria-label': 'Previous week', onclick: () => shiftWeek(-1) },
+        h('button', { class: 'btn btn-icon week-arrow', type: 'button', 'aria-label': 'Previous week', onclick: () => shiftWeek(-1) },
           icon('chevronLeft')),
         h('button', {
           class: 'btn btn-ghost week-label',
@@ -155,12 +160,13 @@ function weekCard() {
           onclick: () => { weekStart = mondayOf(new Date()); refresh(); },
           text: `${dayDate.format(weekStart)} – ${dayDate.format(addDays(weekStart, 6))}`,
         }, h('small', { text: ` Week ${isoWeek(weekStart)}` })),
-        h('button', { class: 'btn btn-ghost btn-icon', type: 'button', 'aria-label': 'Next week', onclick: () => shiftWeek(1) },
-          icon('chevronRight')),
-      ),
-    ),
-    h('ul', { class: 'card days' }, days),
-  ];
+        h('button', { class: 'btn btn-icon week-arrow', type: 'button', 'aria-label': 'Next week', onclick: () => shiftWeek(1) },
+          icon('chevronRight')))),
+    h('ul', { class: 'strip' }, cells),
+    seen.size > 0 && h('div', { class: 'strip-legend' },
+      [...seen].map(([id, name]) => h('span', {},
+        h('i', { class: `strip-dot ${personClass(id)}` }),
+        id === mine ? `${name} (you)` : name))));
 }
 
 // myShifts groups your upcoming dates into runs of consecutive days, then

internal/web/static/js/team.js

@@ -168,13 +168,18 @@ function section() {
 // buttons, because these are six URLs: app.js intercepts the click, the
 // browser's Back walks them, and a reload lands where you were.
 function subnav() {
-  return h('nav', { class: 'subnav', 'aria-label': 'Team' },
+  const nav = h('nav', { class: 'subnav', 'aria-label': 'Team' },
     TABS.map((t) => h('a', {
       class: 'subnav-link',
       href: t.path,
       text: t.label,
       'aria-current': t.tab === tab ? 'page' : null,
     })));
+  // On a phone the strip overflows; bring the open section into view so a
+  // tab past the edge (Sources, Switches) is never the one that is hidden.
+  requestAnimationFrame(() => nav.querySelector('[aria-current]')
+    ?.scrollIntoView({ inline: 'center', block: 'nearest' }));
+  return nav;
 }
 
 // Names which team's settings the six sections below belong to. It used to be
@@ -735,36 +740,19 @@ function minutesInput(seconds, onChange) {
 // --- integrations ----------------------------------------------------------
 
 function integrationsCard() {
-  const rows = (data.integrations || []).map((i) =>
-    h('tr', {},
-      h('td', {}, sourceBadge(i.status)),
-      h('td', { class: 'wrap' },
-        h('strong', { text: i.name }),
-        h('div', { class: 'muted small', text: i.kind })),
-      // When the key last posted, and when an alert last arrived on it. They
-      // differ: a payload with nothing usable in it stamps only the first.
-      h('td', { class: 'muted small' }, timeCell(i.last_used_at)),
-      h('td', { class: 'muted small' }, timeCell(i.last_alert_at)),
-      h('td', { class: 'muted small num', title: 'Distinct alerts refreshed in the last 24 hours',
-        text: String(i.alerts_24h ?? 0) }),
-      h('td', { class: 'muted small' }, h('span', { title: when(i.created_at), text: ago(i.created_at) })),
-      h('td', {}, isOwner() && h('div', { class: 'row-actions' },
-        h('button', {
-          class: 'btn-sm', type: 'button', text: 'Rename', onclick: () => openRenameSource(i),
-        }),
-        h('button', {
-          class: 'btn-sm danger', type: 'button', text: 'Revoke',
-          onclick: async () => {
-            if (!(await confirm({
-              title: `Revoke ${i.name}?`,
-              text: 'Anything posting with this key stops delivering immediately. Alerts it already delivered stay.',
-              confirmLabel: 'Revoke',
-              danger: true,
-            }))) return;
-            act(() => api.deleteIntegration(teamID, i.id));
-          },
-        }))),
-    ));
+  const rows = (data.integrations || []).map((i) => clickableRow(h('tr', {},
+    h('td', {}, sourceBadge(i.status)),
+    h('td', { class: 'wrap' },
+      h('strong', { text: i.name }),
+      h('div', { class: 'muted small', text: i.kind })),
+    // When the key last posted, and when an alert last arrived on it. They
+    // differ: a payload with nothing usable in it stamps only the first.
+    h('td', { class: 'muted small' }, timeCell(i.last_used_at)),
+    h('td', { class: 'muted small' }, timeCell(i.last_alert_at)),
+    h('td', { class: 'muted small num', title: 'Distinct alerts refreshed in the last 24 hours',
+      text: String(i.alerts_24h ?? 0) }),
+    h('td', { class: 'muted small' }, h('span', { title: when(i.created_at), text: ago(i.created_at) })),
+  ), () => openSourceDetail(i)));
 
   return h('div', { class: 'card' },
     h('div', { class: 'card-head' },
@@ -780,9 +768,9 @@ function integrationsCard() {
       ? h('div', { class: 'table-scroll' },
         h('table', { class: 'admin-table status-table' },
           h('thead', {}, h('tr', {},
-            h('th', { text: 'Status' }), h('th', { text: 'Source' }),
+            h('th', { text: 'Status' }), h('th', { text: 'Name' }),
             h('th', { text: 'Last webhook' }), h('th', { text: 'Last alert' }),
-            h('th', { class: 'num', text: 'Alerts 24h' }), h('th', { text: 'Created' }), h('th'))),
+            h('th', { class: 'num', text: 'Alerts 24h' }), h('th', { text: 'Created' }))),
           h('tbody', {}, rows)))
       : h('p', { class: 'muted', text: 'No alert source yet, so nothing can reach this team.' }),
   );
@@ -844,6 +832,34 @@ function openNameSheet({ title, submit, value, run }) {
   openSheet(() => [h('h2', { class: 'sheet-title', text: title }), form]);
 }
 
+function openSourceDetail(i) {
+  openDetailSheet(i.name, [
+    sheetFact('Status', sourceBadge(i.status)),
+    sheetFact('Kind', i.kind),
+    sheetFact('Last webhook', timeCell(i.last_used_at)),
+    sheetFact('Last alert', timeCell(i.last_alert_at)),
+    sheetFact('Alerts, 24h', String(i.alerts_24h ?? 0)),
+    sheetFact('Created', h('span', { title: when(i.created_at), text: ago(i.created_at) })),
+  ],
+  isOwner() && h('button', {
+    class: 'btn', type: 'button', text: 'Rename',
+    onclick: () => { closeSheet(); openRenameSource(i); },
+  }),
+  isOwner() && h('button', {
+    class: 'btn btn-danger', type: 'button', text: 'Revoke',
+    onclick: async () => {
+      closeSheet();
+      if (!(await confirm({
+        title: `Revoke ${i.name}?`,
+        text: 'Anything posting with this key stops delivering immediately. Alerts it already delivered stay.',
+        confirmLabel: 'Revoke',
+        danger: true,
+      }))) return;
+      act(() => api.deleteIntegration(teamID, i.id));
+    },
+  }));
+}
+
 function openNewSource() {
   openNameSheet({
     title: 'New source', submit: 'Add source', value: '',
@@ -899,6 +915,34 @@ const triggeredCell = (iso, incidentID) => {
     : h('span', { title: when(iso), text: ago(iso) });
 };
 
+// A list row that opens its details: the list is for finding the thing, the
+// sheet it opens is where the buttons are. Links inside the row keep working.
+function clickableRow(tr, open) {
+  tr.classList.add('clickable');
+  tr.tabIndex = 0;
+  tr.addEventListener('click', (e) => {
+    if (!e.target.closest('a')) open();
+  });
+  tr.addEventListener('keydown', (e) => {
+    if (e.key === 'Enter' && e.target === tr) open();
+  });
+  return tr;
+}
+
+// Label/value rows for a details sheet, and the sheet itself: facts above, then
+// Close and whatever the viewer may do. Falsy actions (a non-owner's) drop out.
+const sheetFact = (label, value) => [h('dt', { text: label }), h('dd', {}, value)];
+
+function openDetailSheet(title, facts, ...actions) {
+  openSheet(() => [
+    h('h2', { class: 'sheet-title', text: title }),
+    h('dl', { class: 'switch-facts' }, facts),
+    h('div', { class: 'sheet-actions' },
+      h('button', { class: 'btn', type: 'button', text: 'Close', onclick: () => closeSheet() }),
+      ...actions),
+  ]);
+}
+
 function switchRows(sw) {
   const main = h('tr', {},
     h('td', {}, switchBadge(sw.status)),
@@ -908,19 +952,8 @@ function switchRows(sw) {
     h('td', { class: 'muted small' }, timeCell(sw.last_heartbeat_at)),
     h('td', { class: 'muted small' }, triggeredCell(sw.last_triggered_at, sw.open_incident_id)),
     h('td', { class: 'muted small', text: duration(sw.timeout_seconds * 1000) }),
-    h('td', {}, isOwner() && h('button', {
-      class: 'btn-sm danger', type: 'button', text: 'Remove',
-      onclick: async () => {
-        if (!(await confirm({
-          title: `Remove ${sw.name}?`,
-          text: 'It stops being watched. An incident it already opened stays open until it is resolved.',
-          confirmLabel: 'Remove',
-          danger: true,
-        }))) return;
-        act(() => api.deleteDeadmanSwitch(teamID, sw.id));
-      },
-    })),
   );
+  clickableRow(main, () => openSwitchDetail(sw));
 
   // One heartbeat is the switch's own times; several are worth telling apart,
   // since a live cluster must not hide a dead one.
@@ -935,7 +968,7 @@ function switchRows(sw) {
           && h('code', { class: 'small', text: src.fingerprint })),
       h('td', { class: 'muted small' }, timeCell(src.last_heartbeat_at)),
       h('td', { class: 'muted small' }, triggeredCell(src.last_triggered_at, src.incident_id)),
-      h('td'), h('td')))
+      h('td')))
     : [];
   return [main, ...sources];
 }
@@ -947,7 +980,7 @@ function deadmanCard() {
     h('div', { class: 'card-head' },
       h('h2', { text: 'Dead man’s switches' }),
       isOwner() && h('button', {
-        class: 'btn', type: 'button', text: 'New switch', onclick: openNewSwitch,
+        class: 'btn', type: 'button', text: 'New switch', onclick: () => openSwitchForm(),
       })),
     h('p', { class: 'muted small' },
       'Alerts whose ABSENCE is the signal. Receiving one opens nothing; going ',
@@ -956,9 +989,9 @@ function deadmanCard() {
       ? h('div', { class: 'table-scroll' },
         h('table', { class: 'admin-table status-table' },
           h('thead', {}, h('tr', {},
-            h('th', { text: 'Status' }), h('th', { text: 'Switch' }),
+            h('th', { text: 'Status' }), h('th', { text: 'Name' }),
             h('th', { text: 'Last heartbeat' }), h('th', { text: 'Last triggered' }),
-            h('th', { text: 'Silent after' }), h('th'))),
+            h('th', { text: 'Silent after' }))),
           h('tbody', {}, switches.flatMap(switchRows))))
       : h('p', { class: 'muted', text: 'Nothing watched.' }),
   );
@@ -966,16 +999,22 @@ function deadmanCard() {
 
 // The form lives in the sheet, not on the page: most visits are to look at the
 // list, and a form that is always open is the page this replaced.
-function openNewSwitch() {
+function openSwitchForm(existing) {
   const name = h('input', { type: 'text', placeholder: 'Prod Watchdog', autofocus: true });
   const matcher = h('input', {
     type: 'text', placeholder: 'alertname=Watchdog,cluster=prod', class: 'wide', required: true,
   });
   const timeout = h('input', {
-    type: 'number', min: '1', value: '15', class: 'setting-value', required: true,
+    type: 'number', min: '1', step: 'any', value: '15', class: 'setting-value', required: true,
   });
   const severity = h('select', {},
     ...['critical', 'error', 'warning', 'info'].map((s) => h('option', { value: s, text: s })));
+  if (existing) {
+    name.value = existing.name;
+    matcher.value = existing.matcher;
+    timeout.value = String(existing.timeout_seconds / 60);
+    severity.value = existing.severity;
+  }
   const problem = h('p', { class: 'load-error', hidden: true });
 
   const form = h('form', { class: 'stacked-form' },
@@ -990,17 +1029,20 @@ function openNewSwitch() {
     problem,
     h('div', { class: 'sheet-actions' },
       h('button', { class: 'btn', type: 'button', text: 'Cancel', onclick: () => closeSheet(false) }),
-      h('button', { class: 'btn btn-primary', type: 'submit', text: 'Add switch' })));
+      h('button', { class: 'btn btn-primary', type: 'submit', text: existing ? 'Save' : 'Add switch' })));
 
   form.addEventListener('submit', async (e) => {
     e.preventDefault();
     try {
-      await api.createDeadmanSwitch(teamID, {
+      const body = {
         name: name.value.trim(),
         matcher: matcher.value.trim(),
         timeout_seconds: Math.round(Number(timeout.value) * 60),
         severity: severity.value,
-      });
+      };
+      await (existing
+        ? api.updateDeadmanSwitch(teamID, existing.id, body)
+        : api.createDeadmanSwitch(teamID, body));
     } catch (err) {
       problem.textContent = err.message;
       problem.hidden = false;
@@ -1010,7 +1052,47 @@ function openNewSwitch() {
     refresh();
   });
 
-  openSheet(() => [h('h2', { class: 'sheet-title', text: 'New switch' }), form]);
+  openSheet(() => [
+    h('h2', { class: 'sheet-title', text: existing ? 'Edit switch' : 'New switch' }), form]);
+}
+
+// What the list row has no room for, and where Edit and Delete live.
+function openSwitchDetail(sw) {
+  const sources = sw.sources.length > 1
+    ? [sheetFact('Sources', h('div', {}, ...sw.sources.map((src) => h('div', { class: 'source-labels' },
+      switchBadge(src.status),
+      ...Object.entries(src.labels || {})
+        .filter(([k]) => k !== 'alertname')
+        .map(([k, v]) => labelChip(k, v)),
+      h('span', { class: 'muted small' }, ' ', timeCell(src.last_heartbeat_at))))))]
+    : [];
+
+  openDetailSheet(sw.name, [
+    sheetFact('Status', switchBadge(sw.status)),
+    sheetFact('Matcher', h('code', { text: sw.matcher })),
+    sheetFact('Silent after', duration(sw.timeout_seconds * 1000)),
+    sheetFact('Severity', sw.severity),
+    sheetFact('Last heartbeat', timeCell(sw.last_heartbeat_at)),
+    sheetFact('Last triggered', triggeredCell(sw.last_triggered_at, sw.open_incident_id)),
+    sources,
+  ],
+  isOwner() && h('button', {
+    class: 'btn', type: 'button', text: 'Edit',
+    onclick: () => { closeSheet(); openSwitchForm(sw); },
+  }),
+  isOwner() && h('button', {
+    class: 'btn btn-danger', type: 'button', text: 'Delete',
+    onclick: async () => {
+      closeSheet();
+      if (!(await confirm({
+        title: `Remove ${sw.name}?`,
+        text: 'It stops being watched. An incident it already opened stays open until it is resolved.',
+        confirmLabel: 'Remove',
+        danger: true,
+      }))) return;
+      act(() => api.deleteDeadmanSwitch(teamID, sw.id));
+    },
+  }));
 }
 
 // --- members ---------------------------------------------------------------
@@ -1114,45 +1196,18 @@ function membersCard() {
   const members = data.members || [];
   const owners = members.filter((m) => m.role === 'owner').length;
 
-  const rows = members.map((m) => {
-    const lastOwner = m.role === 'owner' && owners === 1;
-    return h('tr', {},
-      h('td', {}, memberBadge(m)),
-      h('td', { class: 'wrap' },
-        h('strong', { text: m.username }),
-        m.user_id === myID() && h('span', { class: 'muted small', text: ' (you)' }),
-        m.problem && h('div', { class: 'target-problem', text: m.problem })),
-      h('td', { class: 'muted small' }, m.role, m.source === 'oidc' && ssoBadge()),
-      h('td', { class: 'muted small' }, shiftCell(m)),
-      h('td', { class: 'muted small' }, timeCell(m.last_active_at)),
-      h('td', { class: 'muted small' },
-        h('span', { title: when(m.joined_at), text: ago(m.joined_at) })),
-      h('td', {}, isOwner() && h('div', { class: 'row-actions' },
-        h('button', {
-          class: 'btn-sm', type: 'button', text: 'Edit',
-          // The server refuses to edit a membership the groups grant.
-          disabled: m.source === 'oidc',
-          title: m.source === 'oidc' ? SSO_MANAGED : null,
-          onclick: () => openEditMember(m),
-        }),
-        h('button', {
-          class: 'btn-sm danger', type: 'button', text: 'Remove',
-          disabled: lastOwner || m.source === 'oidc',
-          title: m.source === 'oidc' ? SSO_MANAGED
-            : lastOwner ? 'A team needs an owner. Make somebody else one first.' : null,
-          onclick: async () => {
-            if (!(await confirm({
-              title: `Remove ${m.username}?`,
-              text: 'They lose access to this team. Rota days already assigned to them are not '
-                + 'changed, so reassign those from the Rota tab.',
-              confirmLabel: 'Remove',
-              danger: true,
-            }))) return;
-            act(() => api.removeTeamMember(teamID, m.user_id));
-          },
-        }))),
-    );
-  });
+  const rows = members.map((m) => clickableRow(h('tr', {},
+    h('td', {}, memberBadge(m)),
+    h('td', { class: 'wrap' },
+      h('strong', { text: m.username }),
+      m.user_id === myID() && h('span', { class: 'muted small', text: ' (you)' }),
+      m.problem && h('div', { class: 'target-problem', text: m.problem })),
+    h('td', { class: 'muted small' }, m.role, m.source === 'oidc' && ssoBadge()),
+    h('td', { class: 'muted small' }, shiftCell(m)),
+    h('td', { class: 'muted small' }, timeCell(m.last_active_at)),
+    h('td', { class: 'muted small' },
+      h('span', { title: when(m.joined_at), text: ago(m.joined_at) })),
+  ), () => openMemberDetail(m, owners)));
 
   return [oidcGroupsCard(), h('div', { class: 'card' },
     h('div', { class: 'card-head' },
@@ -1167,14 +1222,49 @@ function membersCard() {
       ? h('div', { class: 'table-scroll' },
         h('table', { class: 'admin-table status-table' },
           h('thead', {}, h('tr', {},
-            h('th', { text: 'Status' }), h('th', { text: 'Member' }), h('th', { text: 'Role' }),
-            h('th', { text: 'Rota' }), h('th', { text: 'Last active' }), h('th', { text: 'Joined' }),
-            h('th'))),
+            h('th', { text: 'Status' }), h('th', { text: 'Name' }), h('th', { text: 'Role' }),
+            h('th', { text: 'Rota' }), h('th', { text: 'Last active' }), h('th', { text: 'Joined' }))),
           h('tbody', {}, rows)))
       : h('p', { class: 'muted', text: 'Nobody is in this team.' }),
   )];
 }
 
+function openMemberDetail(m, owners) {
+  const lastOwner = m.role === 'owner' && owners === 1;
+  openDetailSheet(m.username, [
+    sheetFact('Status', memberBadge(m)),
+    sheetFact('Role', h('span', {}, m.role, m.source === 'oidc' && ssoBadge())),
+    sheetFact('Rota', shiftCell(m)),
+    sheetFact('Last active', timeCell(m.last_active_at)),
+    sheetFact('Joined', h('span', { title: when(m.joined_at), text: ago(m.joined_at) })),
+    m.problem && sheetFact('Problem', h('span', { class: 'target-problem', text: m.problem })),
+  ],
+  isOwner() && h('button', {
+    class: 'btn', type: 'button', text: 'Edit',
+    // The server refuses to edit a membership the groups grant.
+    disabled: m.source === 'oidc',
+    title: m.source === 'oidc' ? SSO_MANAGED : null,
+    onclick: () => { closeSheet(); openEditMember(m); },
+  }),
+  isOwner() && h('button', {
+    class: 'btn btn-danger', type: 'button', text: 'Remove',
+    disabled: lastOwner || m.source === 'oidc',
+    title: m.source === 'oidc' ? SSO_MANAGED
+      : lastOwner ? 'A team needs an owner. Make somebody else one first.' : null,
+    onclick: async () => {
+      closeSheet();
+      if (!(await confirm({
+        title: `Remove ${m.username}?`,
+        text: 'They lose access to this team. Rota days already assigned to them are not '
+          + 'changed, so reassign those from the Rota tab.',
+        confirmLabel: 'Remove',
+        danger: true,
+      }))) return;
+      act(() => api.removeTeamMember(teamID, m.user_id));
+    },
+  }));
+}
+
 // One sheet for both jobs a member's row has: who, and as what. Adding is
 // choosing a person and a role; editing is the same with the person fixed. The
 // API is one call either way — POST upserts the role.

internal/web/static/js/theme.js

@@ -0,0 +1,45 @@
+// Theme override: System (no attribute, the OS decides), Light or Dark.
+//
+// A classic script loaded from <head>, not a module, so the saved choice is on
+// <html> before the first paint and a Dark user on a light OS never sees a
+// flash. The CSP allows no inline script, hence a file of its own. Account
+// (account.js) reads and writes the choice through window.terdutTheme.
+(function () {
+  var KEY = 'terdut.theme';
+  var COLORS = { light: '#f5f6f8', dark: '#0f1115' };
+
+  function get() {
+    try {
+      var v = localStorage.getItem(KEY);
+      return v === 'light' || v === 'dark' ? v : 'system';
+    } catch (e) {
+      return 'system'; // storage unavailable
+    }
+  }
+
+  // The browser chrome colour follows the choice too; the two <meta>s are
+  // media-keyed to the OS, so a forced theme sets both to the same colour.
+  function apply(theme) {
+    var root = document.documentElement;
+    if (theme === 'system') root.removeAttribute('data-theme');
+    else root.setAttribute('data-theme', theme);
+    var metas = document.querySelectorAll('meta[name="theme-color"]');
+    for (var i = 0; i < metas.length; i++) {
+      var scheme = /dark/.test(metas[i].media) ? 'dark' : 'light';
+      metas[i].content = COLORS[theme === 'system' ? scheme : theme];
+    }
+  }
+
+  function set(theme) {
+    try {
+      if (theme === 'system') localStorage.removeItem(KEY);
+      else localStorage.setItem(KEY, theme);
+    } catch (e) {
+      /* storage unavailable: applies for this page view only */
+    }
+    apply(theme);
+  }
+
+  window.terdutTheme = { get: get, set: set };
+  apply(get());
+})();