feat(server): add scoped chat milestone recovery worker

Reconstruct issue-scoped projection work with the shared coordinator. Signal existing run-status, presentation, and question-receipt transactions; verify real cancellation and continuation producers. Keep production polling until remaining producer coverage is complete.

Co-Authored-By: Paperclip <noreply@paperclip.ing>

doc/plans/2026-10-25-chat-run-milestone-sleep-audit.md

@@ -2,8 +2,8 @@
 
 Baseline: Teams attempt recovery #17592 at f53ae3d6d2.
 Branch: codex/sleep-28-chat-run-milestones.
-Status: projection owners separated and lifecycle cursor regression repaired;
-transaction producer coverage, worker registration and final acceptance pending.
+Status: scoped worker and initial real producer signals implemented and locally
+verified; complete producer coverage, production registration and final acceptance pending.
 
 ## Separate obligations discovered
 
@@ -145,3 +145,99 @@ Paperclip Review result. Neutral is not a completed positive review. #17590 at
 queued at inspection. #17586 also has a neutral Paperclip Review. Final review
 acceptance remains separate. Refreshed master is `2e47c42fad`; stack reconciliation
 remains required. No stack PR was merged.
+
+
+## Scoped worker and initial producer checkpoint
+
+`chat-run-milestone-work.ts` now reuses `registerScopedDeadlineWork`. Its loader
+shares the lifecycle candidate selector with the explicit projector, retaining
+all existing SQL eligibility fences and timestamp precision. It reconstructs
+one entry per company/issue, with a bounded 100-publication projection batch.
+A full batch remains immediately due; a short batch has exhausted the selector
+and disarms, including currently suppressed lineage. Real selection/write
+failures retain the shared retry backoff. The scheduler owns coalescing,
+company fairness, first writes during drain, concurrent changes, and shutdown.
+No new persistent queue or independent timer is introduced.
+
+`notifyChatRunMilestones` carries only company/issue identity through the existing
+outer-transaction protocol. Issue-scoped invalidation includes descendants;
+company scope is available for broader configuration or lifecycle changes.
+Unrelated issues/companies are not reconstructed for an issue hint.
+
+Initial production integrations now await intent before writing:
+
+- `heartbeat/run-lifecycle.ts::writeRunStatus`: both shared status transition
+  callers signal the known company in their existing transaction. This is a
+  conservative company hint; it is not proof that all raw run writes are covered.
+- `heartbeat/run-completion.ts`: the late presentation-decision write now owns
+  a short transaction and signals its issue (company fallback when unbound).
+  Provider execution and presentation generation remain outside that transaction.
+- `question-response-delivery.ts::recordTerminal`: the existing transaction
+  signals the issue before recording the durable target run and outcome.
+
+The production app still uses the combined periodic projector. The new worker
+is exercised through actual coordinator registration in tests but is not yet
+registered by `app.ts`; do not remove the sweep until the remaining producer
+families in the table above are implemented. In particular: queue admission and
+recovery writers, native finalization and semantic receipts, question claim/retry
+transitions, interaction changes, publication settlement, conversation/endpoint
+changes, and decision/wake ledger changes remain to be covered. The late legacy
+presentation signal is implemented, but its own full completion integration
+proof is still required. Current Cloud blockers and single-active-writer limits
+remain unchanged.
+
+### Worker and producer verification
+
+15 real-PostgreSQL worker cases pass in
+`/tmp/sleep28-worker-question-tests.log`. They cover configured idle/internal
+runs/old history, startup and restart, first committed work, outer commit and
+nested rollback, first wake during drain, issue/company isolation, suppressed
+lineage reactivation, late presentation eligibility, 205 notices plus another
+company, selection failure through drain, publication-write retry, a concurrent
+commit during projection, shutdown, real cancellation and real question delivery.
+Quiet assertions advance two minutes and observe zero DB selects/transactions.
+The deadline test verifies no scans during the ten seconds before retry.
+
+The two real producer proofs are causal, not just notification-spy assertions:
+
+- Removing the actual `writeRunStatus` signal makes the cancellation test fail:
+  the queued notice exists but the terminal notice is absent. Restoring it passes.
+  Log: `/tmp/sleep28-cancellation-without-signal.log`.
+- The question test creates and answers a real interaction, supplies a durable
+  published prompt, and invokes the real delivery service. Its wake callback
+  inserts the continuation without a milestone signal and proves no notice is
+  projected yet. The terminal receipt then activates the worker. Removing that
+  exact signal strands the notice and fails the assertion. Log:
+  `/tmp/sleep28-question-without-signal.log`.
+
+The 31-case existing chat integration selection passes after the selector
+refactor and producer additions, including presentation authority and native
+progress. Log: `/tmp/sleep28-scoped-acceptance.log`. Expected contention/retry
+warnings occurred, with all assertions passing. Server production typecheck
+passes in `/tmp/sleep28-producer-typecheck.log`. No full-workspace or PR gate
+completion is claimed yet.
+
+Preserved fixture failures:
+
+1. `/tmp/sleep28-worker-tests.log`: all 11 initial cases failed setup because the
+   fixture used provider name `slack` as connection transport. The schema requires
+   `chat_sdk`; no production transport change was made.
+2. `/tmp/sleep28-worker-fixed-tests.log`: ten passed; the same-company isolation
+   fixture violated the application-name uniqueness constraint. Fixture names now
+   include their identity.
+3. `/tmp/sleep28-worker-producer-tests.log`: eleven passed; cancellation required
+   a responsible-user membership, and teardown omitted its new run-event rows.
+   Both are now represented/cleaned by the fixture.
+4. `/tmp/sleep28-worker-producer-fixed-tests.log`: assertions reached cancellation
+   projection, but default follow-on recovery admitted another execution and
+   teardown deadlocked against it. The test now uses cancellation's existing
+   `suppressImmediateRecovery` option to exercise Stop without starting successor
+   execution. This does not change the default production recovery policy.
+5. All 12 corrected cases passed in `/tmp/sleep28-worker-final-tests.log`; two
+   concurrency/retry cases made 14 pass in `/tmp/sleep28-worker-expanded-tests.log`;
+   the real question case brings the latest total to 15. These counts overlap.
+
+This checkpoint is progress on slice 28, not a new independently complete reason
+or a claim that milestone polling is retired. The original full PR-stack scope,
+older unresolved CI/review findings, final master reconciliation and each PR's
+latest-head acceptance remain required.

server/src/__tests__/chat-run-milestone-work.test.ts

@@ -0,0 +1,261 @@
+import { randomUUID } from "node:crypto";
+import { eq } from "drizzle-orm";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, it, vi } from "vitest";
+import { activityLog, agents, authUsers, companyMemberships, chatConversations, chatEndpoints, chatMessageLinks, chatPublications, companies, createDb,
+  heartbeatRunEvents, heartbeatRuns, issueComments, issueQuestionResponseDeliveries, issueThreadInteractions, issues, toolApplications, toolConnections } from "@paperclipai/db";
+import { getEmbeddedPostgresTestSupport, startEmbeddedPostgresTestDatabase } from "./helpers/embedded-postgres.js";
+import { resolveExternalTestDatabaseUrl } from "./helpers/external-test-database.js";
+import { createDeliveryWorkCoordinator } from "../services/delivery-work-coordinator.js";
+import { chatRunMilestoneWork } from "../services/chat-run-milestone-work.js";
+import { notifyChatRunMilestones } from "../services/chat-run-milestone-notifications.js";
+import { issueThreadInteractionService } from "../services/issue-thread-interactions.js";
+import { questionResponseDeliveryService } from "../services/question-response-delivery.js";
+import { heartbeatService } from "../services/heartbeat.js";
+import { idleWorkSnapshot } from "../services/task-admission.js";
+
+const external = resolveExternalTestDatabaseUrl("PAPERCLIP_MILESTONE_TEST_DATABASE_URL");
+const support = external ? { supported: true } : await getEmbeddedPostgresTestSupport();
+(support.supported ? describe : describe.skip)("chat lifecycle milestone worker", () => {
+  let temporary: Awaited<ReturnType<typeof startEmbeddedPostgresTestDatabase>> | undefined;
+  let db: ReturnType<typeof createDb>;
+  let coordinator: ReturnType<typeof createDeliveryWorkCoordinator> | undefined;
+  let enabled = true;
+  const errors: unknown[] = [];
+  const fixtureCompanies: string[] = [];
+  const fixtureUsers: string[] = [];
+  type Fixture = Awaited<ReturnType<typeof fixture>>;
+  let f: Fixture;
+
+  beforeAll(async () => {
+    if (!external) temporary = await startEmbeddedPostgresTestDatabase("chat-milestones-");
+    db = createDb(external ?? temporary!.connectionString, { maxConnections: 6, idleTimeoutSeconds: 0, maxLifetimeSeconds: 0 });
+  }, 60_000);
+  async function fixture(companyId = randomUUID()) {
+    const agentId = randomUUID(), issueId = randomUUID(), endpointId = randomUUID(), conversationId = randomUUID(), commentId = randomUUID(), userId = randomUUID();
+    if (!fixtureCompanies.includes(companyId)) {
+      fixtureCompanies.push(companyId);
+      await db.insert(companies).values({ id: companyId, name: "Milestones", issuePrefix: `M${companyId.slice(0, 7)}` });
+    }
+    fixtureUsers.push(userId);
+    await db.insert(authUsers).values({ id: userId, name: "Owner", email: `${userId}@example.test`, createdAt: new Date(), updatedAt: new Date() });
+    await db.insert(companyMemberships).values({ companyId, principalType: "user", principalId: userId, status: "active", membershipRole: "owner" });
+    await db.insert(agents).values({ id: agentId, companyId, name: "Maya", adapterType: "paperclip_runner" });
+    await db.insert(issues).values({ id: issueId, companyId, title: "Chat task", status: "in_progress", assigneeAgentId: agentId, responsibleUserId: userId });
+    const applicationId = randomUUID(), connectionId = randomUUID();
+    await db.insert(toolApplications).values({ id: applicationId, companyId, name: `Chat ${applicationId}`, type: "chat" });
+    await db.insert(toolConnections).values({ id: connectionId, companyId, applicationId, name: "Chat", uid: connectionId, transport: "chat_sdk", connectionPurpose: "channel", enabled: true, status: "active" });
+    await db.insert(chatEndpoints).values({ id: endpointId, companyId, connectionId, assignedAgentId: agentId, publicId: endpointId, provider: "slack", status: "active" });
+    await db.insert(chatConversations).values({ id: conversationId, companyId, endpointId, issueId, externalConversationId: conversationId, externalThreadId: conversationId, externalLabel: "Chat" });
+    await db.insert(issueComments).values({ id: commentId, companyId, issueId, authorType: "user", body: "Start work" });
+    await db.insert(chatMessageLinks).values({ companyId, endpointId, conversationId, commentId, providerMessageId: commentId, direction: "inbound" });
+    return { companyId, agentId, issueId, endpointId, conversationId, commentId, userId };
+  }
+  beforeEach(async () => {
+    enabled = true; errors.length = 0; f = await fixture();
+    vi.useFakeTimers({ toFake: ["Date", "setTimeout", "clearTimeout"], shouldAdvanceTime: true });
+  });
+  afterEach(async () => {
+    await coordinator?.stop(); coordinator = undefined; vi.useRealTimers(); vi.restoreAllMocks();
+    for (const companyId of fixtureCompanies.splice(0)) {
+      await db.delete(activityLog).where(eq(activityLog.companyId, companyId));
+      await db.delete(chatMessageLinks).where(eq(chatMessageLinks.companyId, companyId));
+      await db.delete(chatPublications).where(eq(chatPublications.companyId, companyId));
+      await db.delete(chatConversations).where(eq(chatConversations.companyId, companyId));
+      await db.delete(chatEndpoints).where(eq(chatEndpoints.companyId, companyId));
+      await db.delete(issueQuestionResponseDeliveries).where(eq(issueQuestionResponseDeliveries.companyId, companyId));
+      await db.delete(issueThreadInteractions).where(eq(issueThreadInteractions.companyId, companyId));
+      await db.delete(issueComments).where(eq(issueComments.companyId, companyId));
+      await db.delete(heartbeatRunEvents).where(eq(heartbeatRunEvents.companyId, companyId));
+      await db.delete(heartbeatRuns).where(eq(heartbeatRuns.companyId, companyId));
+      await db.delete(issues).where(eq(issues.companyId, companyId));
+      await db.delete(agents).where(eq(agents.companyId, companyId));
+      await db.delete(companyMemberships).where(eq(companyMemberships.companyId, companyId));
+      await db.delete(companies).where(eq(companies.id, companyId));
+    }
+    for (const userId of fixtureUsers.splice(0)) await db.delete(authUsers).where(eq(authUsers.id, userId));
+    expect(errors).toEqual([]);
+  });
+  afterAll(async () => { await db?.$client.end({ timeout: 1 }); await temporary?.cleanup(); });
+  function runValues(scope = f, patch: Partial<typeof heartbeatRuns.$inferInsert> = {}) {
+    return { id: randomUUID(), companyId: scope.companyId, agentId: scope.agentId, responsibleUserId: scope.userId, status: "running", updatedAt: new Date(),
+      contextSnapshot: { issueId: scope.issueId, source: "chat:slack", endpointId: scope.endpointId, wakeCommentId: scope.commentId }, ...patch };
+  }
+  async function run(scope = f, patch: Partial<typeof heartbeatRuns.$inferInsert> = {}) {
+    const values = runValues(scope, patch);
+    await db.transaction(async tx => { await notifyChatRunMilestones(tx, scope); await tx.insert(heartbeatRuns).values(values); });
+    return values;
+  }
+  async function start() {
+    coordinator = createDeliveryWorkCoordinator({ owner: db, canRun: () => enabled, canReconcile: () => true, onError: error => errors.push(error) });
+    await chatRunMilestoneWork(db).register(coordinator, () => enabled).ready;
+  }
+  async function flush() {
+    await vi.waitFor(async () => {
+      await vi.advanceTimersByTimeAsync(1000);
+      expect(idleWorkSnapshot().active).toBe(0); expect(coordinator!.nextWakeAt()).toBeNull();
+    }, { timeout: 10_000 });
+  }
+  async function quiet(ms = 120_000) {
+    const reads = vi.spyOn(db, "select"), writes = vi.spyOn(db, "transaction");
+    try { await vi.advanceTimersByTimeAsync(ms); expect(reads).not.toHaveBeenCalled(); expect(writes).not.toHaveBeenCalled(); }
+    finally { reads.mockRestore(); writes.mockRestore(); }
+  }
+  const publications = (scope = f) => db.select().from(chatPublications).where(eq(chatPublications.issueId, scope.issueId));
+
+  it("leaves configured idle, internal runs and old history quiet", async () => {
+    await run(f, { contextSnapshot: { issueId: f.issueId, source: "issue.comment" } });
+    await run(f, { updatedAt: new Date(Date.now() - 2 * 86_400_000) });
+    await start(); await flush(); expect(await publications()).toEqual([]); await quiet();
+  });
+  it("reconstructs missing notices at startup and leaves published history quiet across restart", async () => {
+    const row = await run(); await start(); await flush();
+    expect(await publications()).toEqual([expect.objectContaining({ idempotencyKey: `run:${row.id}:working:${f.endpointId}` })]);
+    await quiet(); await coordinator!.stop(); await start(); await flush(); await quiet();
+    expect(await publications()).toHaveLength(1);
+  });
+  it("wakes on the first committed row after becoming quiet", async () => {
+    await start(); await flush(); await quiet(); await run(); await flush();
+    expect(await publications()).toHaveLength(1); await quiet();
+  });
+  it("wakes from the real run cancellation transaction without a manual milestone signal", async () => {
+    const row = await run(f, { status: "queued", runtimeMode: "legacy" });
+    await start(); await flush(); await quiet();
+    // This fixture exercises the Stop producer, not successor execution.
+    await heartbeatService(db).cancelRun(row.id, "Operator stopped queued work", { suppressImmediateRecovery: true });
+    await flush();
+    expect((await publications()).map(publication => publication.idempotencyKey).sort()).toEqual([
+      `run:${row.id}:failed:${f.endpointId}`, `run:${row.id}:queued:${f.endpointId}`,
+    ].sort());
+    await quiet();
+  });
+  it("projects a continuation only after the real question delivery receipt commits", async () => {
+    const source = await run(f, { status: "succeeded" });
+    const interactions = issueThreadInteractionService(db);
+    const interaction = await interactions.create({ id: f.issueId, companyId: f.companyId }, {
+      kind: "ask_user_questions", continuationPolicy: "wake_assignee", sourceRunId: source.id,
+      title: "Choose", payload: { version: 1, title: "Choose", questions: [{
+        id: "choice", prompt: "Continue?", selectionMode: "single", required: true,
+        options: [{ id: "yes", label: "Yes" }],
+      }] },
+    }, { agentId: f.agentId, runId: source.id });
+    await interactions.answerQuestions({ id: f.issueId, companyId: f.companyId, status: "in_progress" }, interaction.id,
+      { answers: [{ questionId: "choice", optionIds: ["yes"] }] }, { userId: f.userId });
+    // A provider acknowledgement is durable input to this projection test.
+    await db.insert(chatPublications).values({ companyId: f.companyId, endpointId: f.endpointId, conversationId: f.conversationId,
+      issueId: f.issueId, idempotencyKey: `interaction:${interaction.id}:${f.endpointId}`,
+      payload: { text: "Continue?", interactionId: interaction.id }, state: "published", providerMessageId: "question-1",
+    }).onConflictDoNothing();
+    await db.update(chatPublications).set({ state: "published", providerMessageId: "question-1" }).where(eq(chatPublications.issueId, f.issueId));
+    await start(); await flush(); await quiet();
+    let targetId!: string;
+    const delivery = questionResponseDeliveryService(db, { heartbeat: {
+      cancelRun: async () => { throw new Error("The source is already terminal"); },
+      wakeup: async (_agentId, options) => {
+        const values = runValues(f, { status: "queued", contextSnapshot: options.contextSnapshot });
+        const [target] = await db.insert(heartbeatRuns).values(values).returning();
+        targetId = target.id;
+        await flush();
+        expect((await publications()).some(row => row.idempotencyKey.startsWith(`run:${target.id}:`))).toBe(false);
+        return target;
+      },
+    } });
+    await expect(delivery.deliver(interaction.id)).resolves.toMatchObject({ status: "fallback_queued" });
+    await flush();
+    expect((await publications()).map(row => row.idempotencyKey)).toContain(`run:${targetId}:queued:${f.endpointId}`);
+    await quiet();
+  });
+  it("waits for outer commit and ignores rolled-back nested writes", async () => {
+    await start(); await flush();
+    await expect(db.transaction(async tx => {
+      await tx.transaction(async nested => { await notifyChatRunMilestones(nested, f); await nested.insert(heartbeatRuns).values(runValues()); });
+      throw new Error("rollback fixture");
+    })).rejects.toThrow("rollback fixture");
+    await flush(); expect(await publications()).toEqual([]); await quiet();
+    let release!: () => void, staged!: () => void;
+    const gate = new Promise<void>(resolve => { release = resolve; });
+    const stagedPromise = new Promise<void>(resolve => { staged = resolve; });
+    const writing = db.transaction(async tx => {
+      await tx.transaction(async nested => { await notifyChatRunMilestones(nested, f); await nested.insert(heartbeatRuns).values(runValues()); });
+      staged(); await gate;
+    });
+    try { await stagedPromise; await vi.advanceTimersByTimeAsync(1000); expect(await publications()).toEqual([]); }
+    finally { release(); await writing; }
+    await flush(); expect(await publications()).toHaveLength(1); await quiet();
+  });
+  it("retains the first committed wake during drain without periodic SQL", async () => {
+    await start(); await flush(); enabled = false;
+    await run(); await vi.advanceTimersByTimeAsync(1000);
+    expect(await publications()).toEqual([]); await quiet();
+    enabled = true; await vi.advanceTimersByTimeAsync(30_000); await flush();
+    expect(await publications()).toHaveLength(1); await quiet();
+  });
+  it("scopes mutation wakes to their issue and company", async () => {
+    const sibling = await fixture(f.companyId), foreign = await fixture();
+    await start(); await flush();
+    // Unsupported raw writers deliberately do not wake: they provide a sentinel
+    // proving a supported issue mutation does not perform a global scan.
+    await db.insert(heartbeatRuns).values([runValues(sibling), runValues(foreign)]);
+    await run(); await flush();
+    expect(await publications()).toHaveLength(1); expect(await publications(sibling)).toEqual([]); expect(await publications(foreign)).toEqual([]);
+    await db.transaction(tx => notifyChatRunMilestones(tx, { companyId: f.companyId }));
+    await flush(); expect(await publications(sibling)).toHaveLength(1); expect(await publications(foreign)).toEqual([]);
+    await quiet();
+  });
+  it("disarms suppressed lineage and reactivates after its binding changes", async () => {
+    const row = await run(f, { contextSnapshot: { issueId: f.issueId, source: "chat:slack", endpointId: randomUUID(), wakeCommentId: f.commentId } });
+    await start(); await flush(); expect(await publications()).toEqual([]); await quiet();
+    await db.transaction(async tx => {
+      await notifyChatRunMilestones(tx, f);
+      await tx.update(heartbeatRuns).set({ contextSnapshot: runValues().contextSnapshot }).where(eq(heartbeatRuns.id, row.id));
+    });
+    await flush(); expect(await publications()).toHaveLength(1); await quiet();
+  });
+  it("waits quietly for a late presentation decision", async () => {
+    const row = await run(f, { status: "succeeded" }); await start(); await flush();
+    expect(await publications()).toEqual([]); await quiet();
+    await db.transaction(async tx => {
+      await notifyChatRunMilestones(tx, f);
+      await tx.update(heartbeatRuns).set({ resultJson: { presentationDecision: { source: "none" } } }).where(eq(heartbeatRuns.id, row.id));
+    });
+    await flush(); expect(await publications()).toEqual([expect.objectContaining({ idempotencyKey: `run:${row.id}:completed:${f.endpointId}` })]); await quiet();
+  });
+  it("drains more than one batch without duplicate notices or starving another company", async () => {
+    const other = await fixture();
+    await db.insert(heartbeatRuns).values([...Array.from({ length: 205 }, () => runValues()), runValues(other)]);
+    await start(); await flush(); expect(await publications()).toHaveLength(205); expect(await publications(other)).toHaveLength(1); await quiet();
+  });
+  it("retains a failed selection for retry through drain", async () => {
+    await run();
+    const select = vi.spyOn(db, "select").mockImplementationOnce(() => { throw new Error("temporary read failure"); });
+    await start(); select.mockRestore(); expect(errors.splice(0)).toHaveLength(1);
+    enabled = false; await quiet(); enabled = true; await vi.advanceTimersByTimeAsync(30_000); await flush();
+    expect(await publications()).toHaveLength(1); await quiet();
+  });
+  it("retries a failed publication write without idle scans before its deadline", async () => {
+    await run();
+    const transaction = vi.spyOn(db, "transaction").mockRejectedValueOnce(new Error("temporary publication failure"));
+    await start(); transaction.mockRestore(); expect(errors.splice(0)).toHaveLength(1);
+    expect(await publications()).toEqual([]); await quiet(10_000);
+    await vi.advanceTimersByTimeAsync(30_000); await flush(); expect(await publications()).toHaveLength(1); await quiet();
+  });
+  it("preserves a committed change arriving while a projection batch is running", async () => {
+    await run();
+    const transaction = db.transaction.bind(db);
+    let release!: () => void, entered!: () => void;
+    const gate = new Promise<void>(resolve => { release = resolve; });
+    const staged = new Promise<void>(resolve => { entered = resolve; });
+    const held = vi.spyOn(db, "transaction").mockImplementationOnce(async (callback, config) => {
+      entered(); await gate; return transaction(callback, config);
+    });
+    const starting = start();
+    try {
+      await staged;
+      await transaction(async tx => { await notifyChatRunMilestones(tx, f); await tx.insert(heartbeatRuns).values(runValues()); });
+    } finally { release(); await starting; held.mockRestore(); }
+    await flush(); expect(await publications()).toHaveLength(2); await quiet();
+  });
+  it("stops future wakes and reconstructs writes made after shutdown", async () => {
+    await start(); await flush(); await coordinator!.stop(); await run(); await quiet();
+    expect(await publications()).toEqual([]); await start(); await flush(); expect(await publications()).toHaveLength(1);
+  });
+});

server/src/services/chat-run-milestone-notifications.ts

@@ -0,0 +1,9 @@
+import { DELIVERY_QUEUES, notifyDeliveryWork } from "./delivery-work-notifications.js";
+
+/** Lineage changes can authorize descendants, so invalidate the issue rather
+ * than just the run whose row changed. Configuration may invalidate a company. */
+export type ChatRunMilestoneScope = { companyId: string; issueId?: string };
+
+export function notifyChatRunMilestones(tx: object, scope: ChatRunMilestoneScope) {
+  return notifyDeliveryWork(tx, DELIVERY_QUEUES.chatRunMilestones, JSON.stringify({ companyId: scope.companyId, issueId: scope.issueId }));
+}

server/src/services/chat-run-milestone-work.ts

@@ -0,0 +1,38 @@
+import type { Db } from "@paperclipai/db";
+import type { createDeliveryWorkCoordinator } from "./delivery-work-coordinator.js";
+import { DELIVERY_QUEUES } from "./delivery-work-notifications.js";
+import { registerScopedDeadlineWork } from "./scoped-deadline-work.js";
+import type { ChatRunMilestoneScope } from "./chat-run-milestone-notifications.js";
+import { enqueueChatRunLifecycleMilestones, loadChatRunLifecycleMilestoneScopes } from "./chat-run-publications.js";
+
+type Entry = { id: string; companyId: string; issueId: string; dueAt: number };
+const batchSize = 100;
+
+/** No periodic discovery: reconstruct on startup/committed changes, drain
+ * current notices, and retain a retry only for failure or an unfinished batch. */
+export function chatRunMilestoneWork(db: Db, publicBaseUrl?: string | null) {
+  async function load(scope?: ChatRunMilestoneScope, signal?: AbortSignal): Promise<Entry[]> {
+    return (await loadChatRunLifecycleMilestoneScopes(db, scope, signal)).map(row => ({
+      ...row, id: `${row.companyId}:${row.issueId}`, dueAt: 0,
+    }));
+  }
+  async function check(entry: Entry, signal?: AbortSignal): Promise<Entry | null> {
+    const inserted = await enqueueChatRunLifecycleMilestones(db, {
+      scope: { companyId: entry.companyId, issueId: entry.issueId },
+      publicBaseUrl, limit: batchSize, signal,
+    });
+    // A short batch exhausted the selector, including currently suppressed
+    // candidates. Their dependencies must signal a later eligibility change.
+    return inserted === batchSize ? { ...entry, dueAt: Date.now() } : null;
+  }
+  function register(coordinator: ReturnType<typeof createDeliveryWorkCoordinator>, canRun: () => boolean) {
+    return registerScopedDeadlineWork(coordinator, {
+      queue: DELIVERY_QUEUES.chatRunMilestones,
+      decodeScope: (hint: string): ChatRunMilestoneScope | undefined => hint === "*" ? undefined : JSON.parse(hint),
+      matches: (entry: Entry, scope?: ChatRunMilestoneScope) => !scope ||
+        (entry.companyId === scope.companyId && (!scope.issueId || entry.issueId === scope.issueId)),
+      load, check,
+    }, canRun);
+  }
+  return { load, check, register };
+}

server/src/services/chat-run-publications.ts

@@ -1,3 +1,4 @@
+import type { ChatRunMilestoneScope } from "./chat-run-milestone-notifications.js";
 import { notifyChatPublicationWork } from "./chat-work-notifications.js";
 import {
   and,
@@ -654,21 +655,19 @@ export async function enqueueSafeNativeChatProgress(
   return inserted;
 }
 
-/**
- * Project only coarse run lifecycle into bound external conversations. Raw
- * output, errors, tool events, and reasoning stay in Paperclip. Idempotency is
- * keyed by run, milestone, and endpoint so polling and restarts are harmless.
- */
-export async function enqueueChatRunLifecycleMilestones(
-  db: Db,
-  input: ChatRunProjectionInput = {},
-): Promise<number> {
-  const { since, limit } = chatRunProjectionWindow(input);
-  const issueIdFromContext = boundRunIssueIdSql();
-  const explicitlyAuthoredCommentReason = sql<boolean>`(
+type ChatRunMilestoneCursor = {
+  updatedAt: string;
+  runId: string;
+  conversationId: string;
+};
+
+const explicitlyAuthoredCommentReason = sql<boolean>`(
     ${issueComments.metadata} ->> 'authorizationReason' = 'paperclip_runner_protocol'
     or left(coalesce(${issueComments.metadata} ->> 'authorizationReason', ''), 6) = 'allow_'
   )`;
+
+function lifecycleMilestonePages(db: Db, since: Date, scope?: ChatRunMilestoneScope) {
+  const issueIdFromContext = boundRunIssueIdSql();
   const milestoneFromStatus = sql<string>`case
     when ${heartbeatRuns.status} = 'queued' then 'queued'
     when ${heartbeatRuns.status} = 'running'
@@ -757,21 +756,8 @@ export async function enqueueChatRunLifecycleMilestones(
         and continuation_wake.payload->>'issueId' = ${issueIdFromContext}
     )
   )`;
-  let inserted = 0;
-  let cursor: {
-    updatedAt: string;
-    runId: string;
-    conversationId: string;
-  } | null = null;
-  const bindingsCache = new Map<
-    string,
-    Awaited<ReturnType<typeof resolveChatOriginPublicationBindings>>
-  >();
-  const providerInteractionCache = new Map<string, boolean>();
-  while (inserted < limit) {
-    const pageSize = Math.max(25, Math.min(200, limit - inserted));
-    const pageCursor: typeof cursor = cursor;
-    const rows: ChatRunMilestoneCandidate[] = await db
+  return (pageSize: number, pageCursor: ChatRunMilestoneCursor | null): Promise<ChatRunMilestoneCandidate[]> => {
+    return db
       .select({
         runId: heartbeatRuns.id,
         runStatus: heartbeatRuns.status,
@@ -934,6 +920,8 @@ export async function enqueueChatRunLifecycleMilestones(
                 and target_delivery.status in ('delivered', 'fallback_queued')
             )`,
           ),
+          scope ? eq(heartbeatRuns.companyId, scope.companyId) : undefined,
+          scope?.issueId ? eq(chatConversations.issueId, scope.issueId) : undefined,
           gte(heartbeatRuns.updatedAt, since),
           isNull(chatPublications.id),
           pageCursor
@@ -948,15 +936,59 @@ export async function enqueueChatRunLifecycleMilestones(
         asc(chatConversations.id),
       )
       .limit(pageSize);
+  };
+}
+
+function milestoneCursor(row: ChatRunMilestoneCandidate): ChatRunMilestoneCursor {
+  return { updatedAt: row.runUpdatedAt, runId: row.runId, conversationId: row.conversationId };
+}
+
+/** Reconstruct issue-scoped projection work. Candidate rows are only hints;
+ * projection still applies current lineage and provider authority. */
+export async function loadChatRunLifecycleMilestoneScopes(
+  db: Db,
+  scope?: ChatRunMilestoneScope,
+  signal?: AbortSignal,
+): Promise<Array<{ companyId: string; issueId: string }>> {
+  const pages = lifecycleMilestonePages(db, chatRunProjectionWindow({}).since, scope);
+  const scopes = new Map<string, { companyId: string; issueId: string }>();
+  let cursor: ChatRunMilestoneCursor | null = null;
+  for (;;) {
+    signal?.throwIfAborted();
+    const rows = await pages(100, cursor);
+    for (const row of rows) scopes.set(`${row.companyId}:${row.issueId}`, { companyId: row.companyId, issueId: row.issueId });
+    if (rows.length < 100) return [...scopes.values()];
+    cursor = milestoneCursor(rows[rows.length - 1]);
+  }
+}
+
+/**
+ * Project only coarse run lifecycle into bound external conversations. Raw
+ * output, errors, tool events, and reasoning stay in Paperclip. Idempotency is
+ * keyed by run, milestone, and endpoint so polling and restarts are harmless.
+ */
+export async function enqueueChatRunLifecycleMilestones(
+  db: Db,
+  input: ChatRunProjectionInput & { scope?: ChatRunMilestoneScope; signal?: AbortSignal } = {},
+): Promise<number> {
+  const { since, limit } = chatRunProjectionWindow(input);
+  let inserted = 0;
+  const pages = lifecycleMilestonePages(db, since, input.scope);
+  let cursor: ChatRunMilestoneCursor | null = null;
+  const bindingsCache = new Map<
+    string,
+    Awaited<ReturnType<typeof resolveChatOriginPublicationBindings>>
+  >();
+  const providerInteractionCache = new Map<string, boolean>();
+  while (inserted < limit) {
+    const pageSize = Math.max(25, Math.min(200, limit - inserted));
+    input.signal?.throwIfAborted();
+    const rows = await pages(pageSize, cursor);
     if (rows.length === 0) break;
-    const lastRow = rows.at(-1)!;
-    cursor = {
-      updatedAt: lastRow.runUpdatedAt,
-      runId: lastRow.runId,
-      conversationId: lastRow.conversationId,
-    };
+    cursor = milestoneCursor(rows[rows.length - 1]);
 
     for (const row of rows) {
+      input.signal?.throwIfAborted();
       if (inserted >= limit) break;
       const milestone = milestoneForStatus(row.runStatus, row.runErrorCode);
       if (!milestone) continue;

server/src/services/delivery-work-notifications.ts

@@ -54,6 +54,7 @@ export const DELIVERY_QUEUES = {
   chatVerification: "chat-verification",
   chatInbound: "chat-inbound",
   chatPublications: "chat-publications",
+  chatRunMilestones: "chat-run-milestones",
   chatReceipts: "chat-slack-receipts",
   browser: "browser-use",
   email: "email-channels",

server/src/services/heartbeat/run-completion.ts

@@ -1,3 +1,4 @@
+import { notifyChatRunMilestones } from "../chat-run-milestone-notifications.js";
 import { measureRunPhase } from "../execution-preparation-timing.js";
 import { pendingCaseHandoffForRun } from "./case-handoff-lifecycle.js";
 import { NON_RETRYABLE_PREFLIGHT_FAILURE_CODES, isHeartbeatRunTerminalStatus } from "./run-lifecycle.js";
@@ -889,13 +890,13 @@ export function createHeartbeatRunCompletion(db: Db, dependencies: HeartbeatRunC
           };
         }
 
-        await db
-          .update(heartbeatRuns)
-          .set({
+        await db.transaction(async tx => {
+          await notifyChatRunMilestones(tx, { companyId: livenessRun.companyId, issueId: issueId ?? undefined });
+          await tx.update(heartbeatRuns).set({
             resultJson: sql`coalesce(${heartbeatRuns.resultJson}, '{}'::jsonb) || ${JSON.stringify({ presentationDecision })}::jsonb`,
             updatedAt: new Date(),
-          })
-          .where(eq(heartbeatRuns.id, livenessRun.id));
+          }).where(eq(heartbeatRuns.id, livenessRun.id));
+        });
         await appendRunEvent(livenessRun, {
           eventType: "run.presentation.resolved",
           stream: "system",

server/src/services/heartbeat/run-lifecycle.ts

@@ -1,3 +1,4 @@
+import { notifyChatRunMilestones } from "../chat-run-milestone-notifications.js";
 import { notifyExecutionRunWork } from "../execution-control-notifications.js";
 import os from "node:os";
 import {
@@ -1018,6 +1019,7 @@ export function createHeartbeatLifecycle(db: Db, dependencies: HeartbeatLifecycl
       const tx = transaction as unknown as Db;
       if (isHeartbeatRunTerminalStatus(status)) await notifyDeliveryWork(tx, DELIVERY_QUEUES.accounting, companyId);
       await notifyExecutionRunWork(tx);
+      if (companyId) await notifyChatRunMilestones(tx, { companyId });
       return write(tx);
     });
   }

server/src/services/question-response-delivery.ts

@@ -1,3 +1,4 @@
+import { notifyChatRunMilestones } from "./chat-run-milestone-notifications.js";
 import { and, asc, desc, eq, inArray, isNull, lte, or, sql } from "drizzle-orm";
 import type { Db } from "@paperclipai/db";
 import {
@@ -504,6 +505,7 @@ export function questionResponseDeliveryService(
   }): Promise<QuestionResponseDeliveryOutcome> {
     const at = now();
     const updated = await db.transaction(async (tx) => {
+      await notifyChatRunMilestones(tx, { companyId: input.delivery.companyId, issueId: input.delivery.issueId });
       const row = await tx
         .update(issueQuestionResponseDeliveries)
         .set({