From 9761f18fdfbe62b1ed6f57323cdca98bace7b0d8 Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Fri, 23 Oct 2026 00:10:03 -0700
Subject: [PATCH 1/8] feat: split singleton backend work into opt-in worker
 deployment

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 .github/workflows/tests.yml                   |  3 ++
 .../services/backgroundWorkerStart.test.js    | 34 +++++++++++++
 backend/server.ts                             | 14 ++----
 backend/services/backgroundWorkerStart.ts     | 18 +++++++
 k8s/helm/commonly/templates/_helpers.tpl      | 11 +++++
 .../templates/core/backend-deployment.yaml    | 40 ++++++++++++----
 .../core/backend-worker-deployment.yaml       |  3 ++
 k8s/helm/commonly/values-dev.yaml             | 10 ++++
 k8s/helm/commonly/values.yaml                 |  5 ++
 scripts/verify-backend-worker-chart.sh        | 48 +++++++++++++++++++
 10 files changed, 168 insertions(+), 18 deletions(-)
 create mode 100644 backend/__tests__/unit/services/backgroundWorkerStart.test.js
 create mode 100644 backend/services/backgroundWorkerStart.ts
 create mode 100644 k8s/helm/commonly/templates/core/backend-worker-deployment.yaml
 create mode 100644 scripts/verify-backend-worker-chart.sh

diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
index a32c39cec4..47cbb9d43c 100644
--- a/.github/workflows/tests.yml
+++ b/.github/workflows/tests.yml
@@ -337,6 +337,9 @@ jobs:
             -f k8s/helm/commonly/values-dev.yaml \
             > /tmp/dev-rendered.yaml
 
+      - name: Verify backend worker render
+        run: bash scripts/verify-backend-worker-chart.sh
+
       - name: Lint rendered dev chart
         run: |
           kubeconform --strict --summary -output text \
diff --git a/backend/__tests__/unit/services/backgroundWorkerStart.test.js b/backend/__tests__/unit/services/backgroundWorkerStart.test.js
new file mode 100644
index 0000000000..2415d70b2a
--- /dev/null
+++ b/backend/__tests__/unit/services/backgroundWorkerStart.test.js
@@ -0,0 +1,34 @@
+const { startBackgroundServices } = require('../../../services/backgroundWorkerStart');
+
+describe('background worker start gate', () => {
+  let scheduler;
+  let discordGateway;
+
+  beforeEach(() => {
+    scheduler = { start: jest.fn() };
+    discordGateway = { start: jest.fn() };
+    jest.spyOn(console, 'log').mockImplementation(() => {});
+  });
+
+  afterEach(() => jest.restoreAllMocks());
+
+  test.each([
+    ['unset', undefined, true],
+    ['on', '1', true],
+    ['off', '0', false],
+  ])('RUN_WORKER %s', (_label, runWorker, shouldStart) => {
+    const env = { NODE_ENV: 'production', DISCORD_BOT_TOKEN: 'token' };
+    if (runWorker !== undefined) env.RUN_WORKER = runWorker;
+
+    expect(startBackgroundServices(env, { scheduler, discordGateway })).toBe(shouldStart);
+    expect(scheduler.start).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
+    expect(discordGateway.start).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
+  });
+
+  test('keeps the test environment and tokenless gateway stopped', () => {
+    expect(startBackgroundServices({ NODE_ENV: 'test', DISCORD_BOT_TOKEN: 'token' }, { scheduler, discordGateway })).toBe(false);
+    expect(startBackgroundServices({ NODE_ENV: 'production' }, { scheduler, discordGateway })).toBe(true);
+    expect(scheduler.start).toHaveBeenCalledTimes(1);
+    expect(discordGateway.start).not.toHaveBeenCalled();
+  });
+});
diff --git a/backend/server.ts b/backend/server.ts
index 83d9408c34..32dca671df 100644
--- a/backend/server.ts
+++ b/backend/server.ts
@@ -383,15 +383,11 @@ mongoose.connection.once('open', () => {
 // Start the summarizer scheduler
 const schedulerService = require('./services/schedulerService');
 const discordGatewayService = require('./services/discordGatewayService');
-
-if (process.env.NODE_ENV !== 'test') {
-  console.log('Starting summarizer scheduler...');
-  schedulerService.start();
-
-  if (process.env.DISCORD_BOT_TOKEN) {
-    discordGatewayService.start();
-  }
-}
+const { startBackgroundServices } = require('./services/backgroundWorkerStart');
+startBackgroundServices(process.env, {
+  scheduler: schedulerService,
+  discordGateway: discordGatewayService,
+});
 
 // PostgreSQL status is mounted unconditionally, and NOT as a placeholder:
 // checkStatus reads the pool and the schema itself, so it answers
diff --git a/backend/services/backgroundWorkerStart.ts b/backend/services/backgroundWorkerStart.ts
new file mode 100644
index 0000000000..a567fc43bf
--- /dev/null
+++ b/backend/services/backgroundWorkerStart.ts
@@ -0,0 +1,18 @@
+type BackgroundServices = {
+  scheduler: { start: () => void };
+  discordGateway: { start: () => void };
+};
+
+export function startBackgroundServices(
+  env: NodeJS.ProcessEnv,
+  services: BackgroundServices,
+): boolean {
+  // Unset keeps single-process installs scheduling until the chart moves the
+  // singleton work into its own Deployment.
+  if (env.NODE_ENV === 'test' || env.RUN_WORKER === '0') return false;
+
+  console.log('Starting summarizer scheduler...');
+  services.scheduler.start();
+  if (env.DISCORD_BOT_TOKEN) services.discordGateway.start();
+  return true;
+}
diff --git a/k8s/helm/commonly/templates/_helpers.tpl b/k8s/helm/commonly/templates/_helpers.tpl
index d01b4cd42e..c62bc7bd8d 100644
--- a/k8s/helm/commonly/templates/_helpers.tpl
+++ b/k8s/helm/commonly/templates/_helpers.tpl
@@ -64,6 +64,17 @@ Backend selector labels
 app: backend
 {{- end }}
 
+{{/* The worker must never match the backend Service selector. */}}
+{{- define "commonly.backendWorker.labels" -}}
+{{ include "commonly.labels" . }}
+app: backend-worker
+{{- end }}
+
+{{- define "commonly.backendWorker.selectorLabels" -}}
+{{ include "commonly.selectorLabels" . }}
+app: backend-worker
+{{- end }}
+
 {{/*
 Frontend labels
 */}}
diff --git a/k8s/helm/commonly/templates/core/backend-deployment.yaml b/k8s/helm/commonly/templates/core/backend-deployment.yaml
index 54aa6bbb6a..f93722e661 100644
--- a/k8s/helm/commonly/templates/core/backend-deployment.yaml
+++ b/k8s/helm/commonly/templates/core/backend-deployment.yaml
@@ -1,13 +1,18 @@
-{{- if .Values.backend.enabled }}
+{{- define "commonly.backend.deployment" -}}
 apiVersion: apps/v1
 kind: Deployment
 metadata:
-  name: backend
+  name: {{ if .worker }}backend-worker{{ else }}backend{{ end }}
   namespace: {{ include "commonly.namespace" . }}
   labels:
-    {{- include "commonly.backend.labels" . | nindent 4 }}
+    {{- if .worker }}{{ include "commonly.backendWorker.labels" . | nindent 4 }}{{ else }}{{ include "commonly.backend.labels" . | nindent 4 }}{{ end }}
 spec:
-  replicas: {{ .Values.backend.replicaCount }}
+  replicas: {{ if .worker }}1{{ else }}{{ .Values.backend.replicaCount }}{{ end }}
+  {{- if .worker }}
+  # A rolling surge would briefly run two schedulers and two Discord gateways.
+  strategy:
+    type: Recreate
+  {{- else }}
   # Overlap is safe: the catalog is ephemeral, ObjectStore defaults to Mongo,
   # and request/session state is externalized to Mongo, PostgreSQL, and Redis.
   minReadySeconds: 10
@@ -16,13 +21,14 @@ spec:
     rollingUpdate:
       maxUnavailable: 0
       maxSurge: 1
+  {{- end }}
   selector:
     matchLabels:
-      {{- include "commonly.backend.selectorLabels" . | nindent 6 }}
+      {{- if .worker }}{{ include "commonly.backendWorker.selectorLabels" . | nindent 6 }}{{ else }}{{ include "commonly.backend.selectorLabels" . | nindent 6 }}{{ end }}
   template:
     metadata:
       labels:
-        {{- include "commonly.backend.selectorLabels" . | nindent 8 }}
+        {{- if .worker }}{{ include "commonly.backendWorker.selectorLabels" . | nindent 8 }}{{ else }}{{ include "commonly.backend.selectorLabels" . | nindent 8 }}{{ end }}
     spec:
       serviceAccountName: {{ .Values.rbac.serviceAccount.name }}
       terminationGracePeriodSeconds: 180
@@ -54,7 +60,7 @@ spec:
           mountPath: /catalog
       {{- end }}
       containers:
-      - name: backend
+      - name: {{ if .worker }}backend-worker{{ else }}backend{{ end }}
         image: "{{ .Values.backend.image.repository }}:{{ .Values.backend.image.tag }}"
         imagePullPolicy: {{ .Values.backend.image.pullPolicy }}
         ports:
@@ -64,6 +70,13 @@ spec:
         # Node Environment
         - name: NODE_ENV
           value: {{ .Values.backend.env.nodeEnv }}
+        {{- if .worker }}
+        - name: RUN_WORKER
+          value: "1"
+        {{- else if .Values.backend.worker.enabled }}
+        - name: RUN_WORKER
+          value: "0"
+        {{- end }}
         - name: PORT
           value: {{ .Values.backend.service.port | quote }}
         - name: HOSTED_MCP_HIDDEN_ENTRIES
@@ -693,11 +706,17 @@ spec:
           periodSeconds: {{ .Values.backend.readinessProbe.periodSeconds }}
         {{- end }}
 
-      {{- if and .Values.backend.nodeSelector (gt (len .Values.backend.nodeSelector) 0) }}
+      {{- if and .worker .Values.backend.worker.nodeSelector (gt (len .Values.backend.worker.nodeSelector) 0) }}
+      nodeSelector:
+        {{- toYaml .Values.backend.worker.nodeSelector | nindent 8 }}
+      {{- else if and (not .worker) .Values.backend.nodeSelector (gt (len .Values.backend.nodeSelector) 0) }}
       nodeSelector:
         {{- toYaml .Values.backend.nodeSelector | nindent 8 }}
       {{- end }}
-      {{- if and .Values.backend.tolerations (gt (len .Values.backend.tolerations) 0) }}
+      {{- if and .worker .Values.backend.worker.tolerations (gt (len .Values.backend.worker.tolerations) 0) }}
+      tolerations:
+        {{- toYaml .Values.backend.worker.tolerations | nindent 8 }}
+      {{- else if and (not .worker) .Values.backend.tolerations (gt (len .Values.backend.tolerations) 0) }}
       tolerations:
         {{- toYaml .Values.backend.tolerations | nindent 8 }}
       {{- end }}
@@ -725,4 +744,7 @@ spec:
         configMap:
           name: skills-catalog
       {{- end }}
+{{- end -}}
+{{- if .Values.backend.enabled }}
+{{ include "commonly.backend.deployment" (merge (dict "worker" false) .) }}
 {{- end }}
diff --git a/k8s/helm/commonly/templates/core/backend-worker-deployment.yaml b/k8s/helm/commonly/templates/core/backend-worker-deployment.yaml
new file mode 100644
index 0000000000..93963c33ec
--- /dev/null
+++ b/k8s/helm/commonly/templates/core/backend-worker-deployment.yaml
@@ -0,0 +1,3 @@
+{{- if and .Values.backend.enabled .Values.backend.worker.enabled }}
+{{ include "commonly.backend.deployment" (merge (dict "worker" true) .) }}
+{{- end }}
diff --git a/k8s/helm/commonly/values-dev.yaml b/k8s/helm/commonly/values-dev.yaml
index 843ad1f0d6..187e9ddfdb 100644
--- a/k8s/helm/commonly/values-dev.yaml
+++ b/k8s/helm/commonly/values-dev.yaml
@@ -34,6 +34,16 @@ agentProvisioning:
 
 backend:
   replicaCount: 1
+  worker:
+    enabled: false
+    # Cron and the Discord gateway must survive spot reclamation.
+    nodeSelector:
+      pool: dev
+    tolerations:
+      - key: "pool"
+        operator: "Equal"
+        value: "dev"
+        effect: "NoSchedule"
   image:
     repository: gcr.io/YOUR_GCP_PROJECT_ID/commonly-backend
     tag: "20260417214615-r4"
diff --git a/k8s/helm/commonly/values.yaml b/k8s/helm/commonly/values.yaml
index f1497542c2..0f446a00ab 100644
--- a/k8s/helm/commonly/values.yaml
+++ b/k8s/helm/commonly/values.yaml
@@ -88,6 +88,11 @@ agentProvisioning:
 backend:
   enabled: true
   replicaCount: 1
+  worker:
+    # Opt in only when a dedicated worker can replace web's background jobs.
+    enabled: false
+    nodeSelector: {}
+    tolerations: []
   image:
     repository: gcr.io/YOUR_GCP_PROJECT_ID/commonly-backend
     tag: latest
diff --git a/scripts/verify-backend-worker-chart.sh b/scripts/verify-backend-worker-chart.sh
new file mode 100644
index 0000000000..07241d8e26
--- /dev/null
+++ b/scripts/verify-backend-worker-chart.sh
@@ -0,0 +1,48 @@
+#!/usr/bin/env bash
+set -euo pipefail
+
+chart=k8s/helm/commonly
+yq_bin=${YQ_BIN:-yq}
+tmp_dir=$(mktemp -d)
+trap 'rm -rf "$tmp_dir"' EXIT
+
+for lane in default dev staging; do
+  values=(-f "$chart/values.yaml")
+  if [[ $lane != default ]]; then
+    values+=(-f "$chart/values-dev.yaml")
+  fi
+  if [[ $lane == staging ]]; then
+    values+=(-f "$chart/values-staging.yaml")
+  fi
+
+  helm template "commonly-$lane" "$chart" "${values[@]}" > "$tmp_dir/$lane-off.yaml"
+  helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true > "$tmp_dir/$lane-on.yaml"
+
+  off="$tmp_dir/$lane-off.yaml"
+  on="$tmp_dir/$lane-on.yaml"
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .metadata.name' "$off")" = ""
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .metadata.name' "$on")" = backend-worker
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.replicas' "$on")" = 1
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.replicas' "$on")" = 1
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.strategy.type' "$on")" = Recreate
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$off")" = ""
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$on")" = 0
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$on")" = 1
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.selector.matchLabels.app' "$on")" = backend-worker
+  test "$("$yq_bin" -N 'select(.kind == "Service" and .metadata.name == "backend") | .spec.selector.app' "$on")" = backend
+
+  # With the worker enabled, no existing resource changes except the web
+  # container's RUN_WORKER=0 env. The worker's env otherwise matches web.
+  "$yq_bin" -N '.' "$off" > "$tmp_dir/$lane-normalized-off.yaml"
+  "$yq_bin" -N 'select(.metadata.name != "backend-worker") | del(.spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-normalized-on.yaml"
+  diff -u "$tmp_dir/$lane-normalized-off.yaml" "$tmp_dir/$lane-normalized-on.yaml"
+  "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env | map(select(.name != "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-web-env.yaml"
+  "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env | map(select(.name != "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-worker-env.yaml"
+  diff -u "$tmp_dir/$lane-web-env.yaml" "$tmp_dir/$lane-worker-env.yaml"
+
+  if [[ $lane != default ]]; then
+    test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.nodeSelector.pool' "$on")" = dev
+    test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.tolerations[] | select(.key == "pool" and .value == "dev" and .effect == "NoSchedule") | .key' "$on")" = pool
+    test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env[] | select(.name == "AGENT_PROVISIONER_K8S") | .value' "$on")" = 1
+  fi
+done

From cb070ec18b56d52f684e0f31e89ce39292f7ef25 Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Fri, 23 Oct 2026 00:16:45 -0700
Subject: [PATCH 2/8] test: keep ops-runner env mirror scoped to shared
 dependencies

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 .../__tests__/unit/helm/opsRunnerEnvMirror.test.js    | 11 +++++++++--
 .../commonly/templates/core/backend-deployment.yaml   |  2 +-
 .../commonly/templates/ops/ops-runner-deployment.yaml |  5 +++--
 3 files changed, 13 insertions(+), 5 deletions(-)

diff --git a/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js b/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js
index 16b4fc3f87..d2c3524f6f 100644
--- a/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js
+++ b/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js
@@ -6,7 +6,8 @@
  * backend's byte-for-byte render — the env block has blank lines nindent would
  * re-indent). A "KEEP IN SYNC" banner alone drifts (recorded instances across
  * the chart), so this pins the contract: the two Deployments must declare the
- * SAME env entries — the same (name, source) tuples, where source is a
+ * SAME env entries except RUN_WORKER, which controls the serving process and
+ * is not read by the sleeping ops-runner — the same (name, source) tuples, where source is a
  * secretKeyRef name+key, a configMapKeyRef name+key, or a literal value. If
  * backend gains/loses an env var, or changes a var's source, and ops-runner
  * isn't updated to match, this fails.
@@ -93,7 +94,13 @@ if (typeof describe === 'function') {
     // ops-runner the same number of times, and vice versa. toEqual on the
     // sorted arrays reports exactly which env entry drifted.
     test('ops-runner declares the same env (name + source) tuples as backend', () => {
-      expect([...opsTuples].sort()).toEqual([...backendTuples].sort());
+      const withoutWorkerGate = (tuple) => !tuple.startsWith('RUN_WORKER\t');
+      expect([...opsTuples].filter(withoutWorkerGate).sort())
+        .toEqual([...backendTuples].filter(withoutWorkerGate).sort());
+      expect(backendTuples.filter((tuple) => !withoutWorkerGate(tuple))).toEqual([
+        'RUN_WORKER\tvalue:"1"',
+        'RUN_WORKER\tvalue:"0"',
+      ]);
     });
   });
 }
diff --git a/k8s/helm/commonly/templates/core/backend-deployment.yaml b/k8s/helm/commonly/templates/core/backend-deployment.yaml
index f93722e661..ee471e4981 100644
--- a/k8s/helm/commonly/templates/core/backend-deployment.yaml
+++ b/k8s/helm/commonly/templates/core/backend-deployment.yaml
@@ -60,7 +60,7 @@ spec:
           mountPath: /catalog
       {{- end }}
       containers:
-      - name: {{ if .worker }}backend-worker{{ else }}backend{{ end }}
+      - name: backend
         image: "{{ .Values.backend.image.repository }}:{{ .Values.backend.image.tag }}"
         imagePullPolicy: {{ .Values.backend.image.pullPolicy }}
         ports:
diff --git a/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml b/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
index 98b89cb749..85b1843753 100644
--- a/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
+++ b/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
@@ -33,8 +33,9 @@ spec:
         # sleep so the container stays up as an exec target; it serves no traffic.
         command: ["/bin/sh", "-c", "sleep infinity"]
         # ===================================================================
-        # ENV + VOLUME MOUNTS below are MIRRORED VERBATIM from
-        # templates/core/backend-deployment.yaml -- KEEP IN SYNC WITH IT.
+        # ENV + VOLUME MOUNTS below mirror backend's application dependencies
+        # from templates/core/backend-deployment.yaml -- KEEP IN SYNC WITH IT.
+        # RUN_WORKER is the server start gate, not an ops-runner dependency.
         # A DRY helper extraction can't keep backend's render byte-identical:
         # the env block carries blank lines that `nindent` would re-indent.
         # Operator scripts here see the same MONGO_URI / PG_* / JWT_SECRET /

From 1dce38af1a08c4e6e56c598bbb75782e437a28fb Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Fri, 23 Oct 2026 00:18:42 -0700
Subject: [PATCH 3/8] docs: clarify ops-runner excludes worker start flag

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js    | 8 +++++---
 .../commonly/templates/ops/ops-runner-deployment.yaml     | 3 ++-
 2 files changed, 7 insertions(+), 4 deletions(-)

diff --git a/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js b/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js
index d2c3524f6f..cae7e56132 100644
--- a/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js
+++ b/backend/__tests__/unit/helm/opsRunnerEnvMirror.test.js
@@ -1,13 +1,14 @@
 /**
  * Drift guard for the ops-runner env mirror.
  *
- * templates/ops/ops-runner-deployment.yaml mirrors the backend container's env
- * + volume mounts VERBATIM (it can't DRY-extract into a helper without changing
+ * templates/ops/ops-runner-deployment.yaml mirrors the backend container's app
+ * dependencies and volume mounts (it can't DRY-extract into a helper without changing
  * backend's byte-for-byte render — the env block has blank lines nindent would
  * re-indent). A "KEEP IN SYNC" banner alone drifts (recorded instances across
  * the chart), so this pins the contract: the two Deployments must declare the
  * SAME env entries except RUN_WORKER, which controls the serving process and
- * is not read by the sleeping ops-runner — the same (name, source) tuples, where source is a
+ * is not read by the sleeping ops-runner — the same (name, source) tuples,
+ * where source is a
  * secretKeyRef name+key, a configMapKeyRef name+key, or a literal value. If
  * backend gains/loses an env var, or changes a var's source, and ops-runner
  * isn't updated to match, this fails.
@@ -97,6 +98,7 @@ if (typeof describe === 'function') {
       const withoutWorkerGate = (tuple) => !tuple.startsWith('RUN_WORKER\t');
       expect([...opsTuples].filter(withoutWorkerGate).sort())
         .toEqual([...backendTuples].filter(withoutWorkerGate).sort());
+      expect(opsTuples.every(withoutWorkerGate)).toBe(true);
       expect(backendTuples.filter((tuple) => !withoutWorkerGate(tuple))).toEqual([
         'RUN_WORKER\tvalue:"1"',
         'RUN_WORKER\tvalue:"0"',
diff --git a/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml b/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
index 85b1843753..a934f7649f 100644
--- a/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
+++ b/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
@@ -4,7 +4,8 @@
 # cgroup/CPU limit and starve its readiness probe (2026-10-22 11:42Z incident;
 # see docs/runbooks/ops-runner-and-no-exec-into-backend.md). Same image + env +
 # secret mounts as backend; replicas 1, spot-pool, no Service, no ingress, no
-# probes, `sleep infinity`. Use: kubectl exec -i deploy/ops-runner --
+# probes, `sleep infinity`. RUN_WORKER is reserved for the serving process.
+# Use: kubectl exec -i deploy/ops-runner --
 # env NODE_PATH=/app/node_modules node - < your-script.js
 apiVersion: apps/v1
 kind: Deployment

From 42dbd95c9b6f435069ba1c17327e764207248b0e Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Fri, 23 Oct 2026 00:31:29 -0700
Subject: [PATCH 4/8] fix: stage worker activation without overlapping web jobs

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 .github/workflows/tests.yml                               | 7 ++++++-
 k8s/helm/commonly/templates/core/backend-deployment.yaml  | 2 +-
 .../commonly/templates/ops/ops-runner-deployment.yaml     | 8 +++-----
 k8s/helm/commonly/values.yaml                             | 3 +++
 scripts/verify-backend-worker-chart.sh                    | 7 +++++++
 5 files changed, 20 insertions(+), 7 deletions(-)

diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
index 47cbb9d43c..c07866fb8f 100644
--- a/.github/workflows/tests.yml
+++ b/.github/workflows/tests.yml
@@ -336,6 +336,11 @@ jobs:
             -f k8s/helm/commonly/values.yaml \
             -f k8s/helm/commonly/values-dev.yaml \
             > /tmp/dev-rendered.yaml
+          helm template commonly-dev-worker k8s/helm/commonly \
+            -f k8s/helm/commonly/values.yaml \
+            -f k8s/helm/commonly/values-dev.yaml \
+            --set backend.worker.enabled=true \
+            > /tmp/dev-worker-rendered.yaml
 
       - name: Verify backend worker render
         run: bash scripts/verify-backend-worker-chart.sh
@@ -345,7 +350,7 @@ jobs:
           kubeconform --strict --summary -output text \
             --schema-location "https://raw.githubusercontent.com/datreeio/CRDs-catalog/${CRDS_CATALOG_SHA}/{{.Group}}/{{.ResourceKind}}_{{.ResourceAPIVersion}}.json" \
             --schema-location default \
-            /tmp/dev-rendered.yaml
+            /tmp/dev-rendered.yaml /tmp/dev-worker-rendered.yaml
 
       # The dev render has kubeconform coverage, but staging is otherwise first
       # rendered at deploy time. Parse both values combinations before merge.
diff --git a/k8s/helm/commonly/templates/core/backend-deployment.yaml b/k8s/helm/commonly/templates/core/backend-deployment.yaml
index ee471e4981..61517210b9 100644
--- a/k8s/helm/commonly/templates/core/backend-deployment.yaml
+++ b/k8s/helm/commonly/templates/core/backend-deployment.yaml
@@ -7,7 +7,7 @@ metadata:
   labels:
     {{- if .worker }}{{ include "commonly.backendWorker.labels" . | nindent 4 }}{{ else }}{{ include "commonly.backend.labels" . | nindent 4 }}{{ end }}
 spec:
-  replicas: {{ if .worker }}1{{ else }}{{ .Values.backend.replicaCount }}{{ end }}
+  replicas: {{ if .worker }}{{ if .Values.backend.worker.startPaused }}0{{ else }}1{{ end }}{{ else }}{{ .Values.backend.replicaCount }}{{ end }}
   {{- if .worker }}
   # A rolling surge would briefly run two schedulers and two Discord gateways.
   strategy:
diff --git a/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml b/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
index a934f7649f..98b89cb749 100644
--- a/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
+++ b/k8s/helm/commonly/templates/ops/ops-runner-deployment.yaml
@@ -4,8 +4,7 @@
 # cgroup/CPU limit and starve its readiness probe (2026-10-22 11:42Z incident;
 # see docs/runbooks/ops-runner-and-no-exec-into-backend.md). Same image + env +
 # secret mounts as backend; replicas 1, spot-pool, no Service, no ingress, no
-# probes, `sleep infinity`. RUN_WORKER is reserved for the serving process.
-# Use: kubectl exec -i deploy/ops-runner --
+# probes, `sleep infinity`. Use: kubectl exec -i deploy/ops-runner --
 # env NODE_PATH=/app/node_modules node - < your-script.js
 apiVersion: apps/v1
 kind: Deployment
@@ -34,9 +33,8 @@ spec:
         # sleep so the container stays up as an exec target; it serves no traffic.
         command: ["/bin/sh", "-c", "sleep infinity"]
         # ===================================================================
-        # ENV + VOLUME MOUNTS below mirror backend's application dependencies
-        # from templates/core/backend-deployment.yaml -- KEEP IN SYNC WITH IT.
-        # RUN_WORKER is the server start gate, not an ops-runner dependency.
+        # ENV + VOLUME MOUNTS below are MIRRORED VERBATIM from
+        # templates/core/backend-deployment.yaml -- KEEP IN SYNC WITH IT.
         # A DRY helper extraction can't keep backend's render byte-identical:
         # the env block carries blank lines that `nindent` would re-indent.
         # Operator scripts here see the same MONGO_URI / PG_* / JWT_SECRET /
diff --git a/k8s/helm/commonly/values.yaml b/k8s/helm/commonly/values.yaml
index 0f446a00ab..b9b86cc3a2 100644
--- a/k8s/helm/commonly/values.yaml
+++ b/k8s/helm/commonly/values.yaml
@@ -91,6 +91,9 @@ backend:
   worker:
     # Opt in only when a dedicated worker can replace web's background jobs.
     enabled: false
+    # First enable with this true; wait for old web pods to terminate, then
+    # clear it so the worker starts without overlapping the old scheduler.
+    startPaused: false
     nodeSelector: {}
     tolerations: []
   image:
diff --git a/scripts/verify-backend-worker-chart.sh b/scripts/verify-backend-worker-chart.sh
index 07241d8e26..a709344ba6 100644
--- a/scripts/verify-backend-worker-chart.sh
+++ b/scripts/verify-backend-worker-chart.sh
@@ -17,13 +17,17 @@ for lane in default dev staging; do
 
   helm template "commonly-$lane" "$chart" "${values[@]}" > "$tmp_dir/$lane-off.yaml"
   helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true > "$tmp_dir/$lane-on.yaml"
+  helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true --set backend.worker.startPaused=true > "$tmp_dir/$lane-paused.yaml"
 
   off="$tmp_dir/$lane-off.yaml"
   on="$tmp_dir/$lane-on.yaml"
+  paused="$tmp_dir/$lane-paused.yaml"
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .metadata.name' "$off")" = ""
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .metadata.name' "$on")" = backend-worker
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.replicas' "$on")" = 1
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.replicas' "$on")" = 1
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.replicas' "$paused")" = 0
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$paused")" = 0
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.strategy.type' "$on")" = Recreate
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$off")" = ""
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$on")" = 0
@@ -39,6 +43,9 @@ for lane in default dev staging; do
   "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env | map(select(.name != "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-web-env.yaml"
   "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env | map(select(.name != "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-worker-env.yaml"
   diff -u "$tmp_dir/$lane-web-env.yaml" "$tmp_dir/$lane-worker-env.yaml"
+  "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].image' "$on" > "$tmp_dir/$lane-web-image.yaml"
+  "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].image' "$on" > "$tmp_dir/$lane-worker-image.yaml"
+  diff -u "$tmp_dir/$lane-web-image.yaml" "$tmp_dir/$lane-worker-image.yaml"
 
   if [[ $lane != default ]]; then
     test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.nodeSelector.pool' "$on")" = dev

From b801795e5db12a8821939994c1aa35306472caa8 Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Fri, 23 Oct 2026 00:39:23 -0700
Subject: [PATCH 5/8] fix: default new worker to paused during handoff

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 .github/workflows/tests.yml            | 1 +
 k8s/helm/commonly/values.yaml          | 6 +++---
 scripts/verify-backend-worker-chart.sh | 4 ++--
 3 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
index c07866fb8f..a6e01dee17 100644
--- a/.github/workflows/tests.yml
+++ b/.github/workflows/tests.yml
@@ -340,6 +340,7 @@ jobs:
             -f k8s/helm/commonly/values.yaml \
             -f k8s/helm/commonly/values-dev.yaml \
             --set backend.worker.enabled=true \
+            --set backend.worker.startPaused=false \
             > /tmp/dev-worker-rendered.yaml
 
       - name: Verify backend worker render
diff --git a/k8s/helm/commonly/values.yaml b/k8s/helm/commonly/values.yaml
index b9b86cc3a2..0710666d97 100644
--- a/k8s/helm/commonly/values.yaml
+++ b/k8s/helm/commonly/values.yaml
@@ -91,9 +91,9 @@ backend:
   worker:
     # Opt in only when a dedicated worker can replace web's background jobs.
     enabled: false
-    # First enable with this true; wait for old web pods to terminate, then
-    # clear it so the worker starts without overlapping the old scheduler.
-    startPaused: false
+    # Enabling the worker first disables web jobs but keeps worker replicas at
+    # zero. After old web pods terminate, set this false to start the worker.
+    startPaused: true
     nodeSelector: {}
     tolerations: []
   image:
diff --git a/scripts/verify-backend-worker-chart.sh b/scripts/verify-backend-worker-chart.sh
index a709344ba6..beac6e1efa 100644
--- a/scripts/verify-backend-worker-chart.sh
+++ b/scripts/verify-backend-worker-chart.sh
@@ -16,8 +16,8 @@ for lane in default dev staging; do
   fi
 
   helm template "commonly-$lane" "$chart" "${values[@]}" > "$tmp_dir/$lane-off.yaml"
-  helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true > "$tmp_dir/$lane-on.yaml"
-  helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true --set backend.worker.startPaused=true > "$tmp_dir/$lane-paused.yaml"
+  helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true --set backend.worker.startPaused=false > "$tmp_dir/$lane-on.yaml"
+  helm template "commonly-$lane" "$chart" "${values[@]}" --set backend.worker.enabled=true > "$tmp_dir/$lane-paused.yaml"
 
   off="$tmp_dir/$lane-off.yaml"
   on="$tmp_dir/$lane-on.yaml"

From a0a99a94285993c1a06ae82f04c52d950ffc358c Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Fri, 23 Oct 2026 00:48:53 -0700
Subject: [PATCH 6/8] test: keep non-worker resources visible in render diff

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 scripts/verify-backend-worker-chart.sh | 5 ++++-
 1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/scripts/verify-backend-worker-chart.sh b/scripts/verify-backend-worker-chart.sh
index beac6e1efa..b603edee27 100644
--- a/scripts/verify-backend-worker-chart.sh
+++ b/scripts/verify-backend-worker-chart.sh
@@ -27,18 +27,21 @@ for lane in default dev staging; do
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.replicas' "$on")" = 1
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.replicas' "$on")" = 1
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.replicas' "$paused")" = 0
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.replicas' "$paused")" = 1
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$paused")" = 0
+  test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$paused")" = 1
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.strategy.type' "$on")" = Recreate
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$off")" = ""
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$on")" = 0
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER") | .value' "$on")" = 1
   test "$("$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.selector.matchLabels.app' "$on")" = backend-worker
   test "$("$yq_bin" -N 'select(.kind == "Service" and .metadata.name == "backend") | .spec.selector.app' "$on")" = backend
+  test "$("$yq_bin" -N 'select((.kind == "Service" or .kind == "Ingress") and .metadata.name == "backend-worker") | .kind' "$on")" = ""
 
   # With the worker enabled, no existing resource changes except the web
   # container's RUN_WORKER=0 env. The worker's env otherwise matches web.
   "$yq_bin" -N '.' "$off" > "$tmp_dir/$lane-normalized-off.yaml"
-  "$yq_bin" -N 'select(.metadata.name != "backend-worker") | del(.spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-normalized-on.yaml"
+  "$yq_bin" -N 'select(.kind != "Deployment" or .metadata.name != "backend-worker") | del(.spec.template.spec.containers[].env[] | select(.name == "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-normalized-on.yaml"
   diff -u "$tmp_dir/$lane-normalized-off.yaml" "$tmp_dir/$lane-normalized-on.yaml"
   "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend") | .spec.template.spec.containers[].env | map(select(.name != "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-web-env.yaml"
   "$yq_bin" -N 'select(.kind == "Deployment" and .metadata.name == "backend-worker") | .spec.template.spec.containers[].env | map(select(.name != "RUN_WORKER"))' "$on" > "$tmp_dir/$lane-worker-env.yaml"

From c2dfd5a221f6e71e866954c32879a98919ade8b4 Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Sat, 24 Oct 2026 00:18:27 -0700
Subject: [PATCH 7/8] fix: gate PG cron startup to backend worker

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6-sol/medium
Agent-Model: codex/gpt-6.1-sol/medium
---
 backend/__tests__/unit/server.test.js     | 41 +++++++++++++++++++++++
 backend/server.ts                         |  4 +--
 backend/services/backgroundWorkerStart.ts | 10 ++++--
 3 files changed, 50 insertions(+), 5 deletions(-)

diff --git a/backend/__tests__/unit/server.test.js b/backend/__tests__/unit/server.test.js
index 0e56844417..a1a9c1ee5e 100644
--- a/backend/__tests__/unit/server.test.js
+++ b/backend/__tests__/unit/server.test.js
@@ -48,6 +48,18 @@ jest.mock('../../config/init-pg-db', () => {
   const initPGDB = mockInitPGDB;
   return initPGDB;
 });
+let mockInitPgRetention = jest.fn();
+jest.mock('../../services/pgRetentionService', () => {
+  const initPgRetention = mockInitPgRetention;
+  return { initPgRetention };
+});
+let mockInitInstallationCleanup = jest.fn();
+jest.mock('../../services/agentInstallationCleanupService', () => {
+  const initInstallationCleanup = mockInitInstallationCleanup;
+  return { initInstallationCleanup };
+});
+jest.mock('../../services/schedulerService', () => ({ start: jest.fn() }));
+jest.mock('../../services/discordGatewayService', () => ({ start: jest.fn() }));
 
 // Replace pg routes with simple routers
 jest.mock('../../routes/pg-status', () => {
@@ -89,6 +101,8 @@ describe('server pg boot routes', () => {
   beforeEach(() => {
     mockConnectPG = jest.fn().mockResolvedValue(null);
     mockInitPGDB = jest.fn();
+    mockInitPgRetention = jest.fn();
+    mockInitInstallationCleanup = jest.fn();
   });
 
   afterEach(() => {
@@ -96,6 +110,7 @@ describe('server pg boot routes', () => {
     jest.clearAllMocks();
     delete process.env.PG_HOST;
     delete process.env.PG_BOOT_RETRY_BASE_DELAY_MS;
+    delete process.env.RUN_WORKER;
   });
 
   it('mounts the real status router even when PG is not configured', async () => {
@@ -161,6 +176,32 @@ describe('server pg boot routes', () => {
     await request(app).get('/api/pg/messages').expect(404);
   });
 
+  it.each([
+    ['web', '0', false],
+    ['worker', '1', true],
+    ['legacy unset', undefined, true],
+  ])('%s boot mounts PG routes and applies the worker cron gate', async (_label, runWorker, shouldStart) => {
+    const priorNodeEnv = process.env.NODE_ENV;
+    process.env.NODE_ENV = 'production';
+    process.env.PG_HOST = 'x';
+    if (runWorker !== undefined) process.env.RUN_WORKER = runWorker;
+    mockConnectPG.mockResolvedValue({});
+    mockInitPGDB.mockResolvedValue(true);
+    try {
+      // This imports the actual server boot composition, including its
+      // createPgBoot onMounted callback. Only the cron starters are mocked.
+      // eslint-disable-next-line global-require, import/no-unresolved, import/extensions
+      const { app } = require('../../server');
+      await new Promise((resolve) => { setTimeout(resolve, 30); });
+      await request(app).get('/api/pg/messages').expect(200);
+      expect(mockInitPgRetention).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
+      expect(mockInitInstallationCleanup).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
+    } finally {
+      if (priorNodeEnv === undefined) delete process.env.NODE_ENV;
+      else process.env.NODE_ENV = priorNodeEnv;
+    }
+  });
+
   it('wires the readiness probe to the route table, not to the boot flag (TASK-168)', () => {
     // The states reachable in this file cannot tell the two apart — when PG is
     // missing, the flag and the route table are both false; when it works, both
diff --git a/backend/server.ts b/backend/server.ts
index 32dca671df..c22f31b18f 100644
--- a/backend/server.ts
+++ b/backend/server.ts
@@ -383,7 +383,7 @@ mongoose.connection.once('open', () => {
 // Start the summarizer scheduler
 const schedulerService = require('./services/schedulerService');
 const discordGatewayService = require('./services/discordGatewayService');
-const { startBackgroundServices } = require('./services/backgroundWorkerStart');
+const { startBackgroundServices, shouldRunBackgroundServices } = require('./services/backgroundWorkerStart');
 startBackgroundServices(process.env, {
   scheduler: schedulerService,
   discordGateway: discordGatewayService,
@@ -437,7 +437,7 @@ if (process.env.PG_HOST) {
       // Kick off the daily 30-day message retention cron. Kept out
       // of schedulerService.ts on purpose so other tracks can edit
       // that file without stomping on this cron.
-      if (process.env.NODE_ENV !== 'test') {
+      if (shouldRunBackgroundServices(process.env)) {
         try {
           const { initPgRetention } = require('./services/pgRetentionService');
           initPgRetention();
diff --git a/backend/services/backgroundWorkerStart.ts b/backend/services/backgroundWorkerStart.ts
index a567fc43bf..5f59cf18ec 100644
--- a/backend/services/backgroundWorkerStart.ts
+++ b/backend/services/backgroundWorkerStart.ts
@@ -7,12 +7,16 @@ export function startBackgroundServices(
   env: NodeJS.ProcessEnv,
   services: BackgroundServices,
 ): boolean {
-  // Unset keeps single-process installs scheduling until the chart moves the
-  // singleton work into its own Deployment.
-  if (env.NODE_ENV === 'test' || env.RUN_WORKER === '0') return false;
+  if (!shouldRunBackgroundServices(env)) return false;
 
   console.log('Starting summarizer scheduler...');
   services.scheduler.start();
   if (env.DISCORD_BOT_TOKEN) services.discordGateway.start();
   return true;
 }
+
+export function shouldRunBackgroundServices(env: NodeJS.ProcessEnv): boolean {
+  // Unset keeps single-process installs scheduling until the chart moves the
+  // singleton work into its own Deployment.
+  return env.NODE_ENV !== 'test' && env.RUN_WORKER !== '0';
+}

From 9b72feaf49769ee0f02f9873ca00ad469e893976 Mon Sep 17 00:00:00 2001
From: Lily Shen <115414357+lilyshen0722@users.noreply.github.com>
Date: Sun, 25 Oct 2026 04:57:39 -0700
Subject: [PATCH 8/8] test(worker): cover reconciler cron in rebased server
 boot

Co-authored-by: flint <WyJmbGludCIsImRlZmF1bHQiXQ@agents.commonly.invalid>
Agent-Model: codex/gpt-6.1-sol/medium
---
 backend/__tests__/unit/server.test.js | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/backend/__tests__/unit/server.test.js b/backend/__tests__/unit/server.test.js
index a1a9c1ee5e..1201869022 100644
--- a/backend/__tests__/unit/server.test.js
+++ b/backend/__tests__/unit/server.test.js
@@ -58,6 +58,11 @@ jest.mock('../../services/agentInstallationCleanupService', () => {
   const initInstallationCleanup = mockInitInstallationCleanup;
   return { initInstallationCleanup };
 });
+let mockInitPgMessageReconcile = jest.fn();
+jest.mock('../../services/pgMessageReconcileService', () => {
+  const initPgMessageReconcile = mockInitPgMessageReconcile;
+  return { initPgMessageReconcile };
+});
 jest.mock('../../services/schedulerService', () => ({ start: jest.fn() }));
 jest.mock('../../services/discordGatewayService', () => ({ start: jest.fn() }));
 
@@ -103,6 +108,7 @@ describe('server pg boot routes', () => {
     mockInitPGDB = jest.fn();
     mockInitPgRetention = jest.fn();
     mockInitInstallationCleanup = jest.fn();
+    mockInitPgMessageReconcile = jest.fn();
   });
 
   afterEach(() => {
@@ -196,6 +202,7 @@ describe('server pg boot routes', () => {
       await request(app).get('/api/pg/messages').expect(200);
       expect(mockInitPgRetention).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
       expect(mockInitInstallationCleanup).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
+      expect(mockInitPgMessageReconcile).toHaveBeenCalledTimes(shouldStart ? 1 : 0);
     } finally {
       if (priorNodeEnv === undefined) delete process.env.NODE_ENV;
       else process.env.NODE_ENV = priorNodeEnv;