Codex review: needs maintainer review before merge. Reviewed May 27, 2026, 7:06 PM ET / 23:06 UTC.

Summary
The branch removes heartbeat direct replay of non-ack pending final text, keeps ack cleanup, adds an isolated active-run guard, and updates focused heartbeat/reply tests.

PR surface: Source -20, Tests +104. Total +84 across 5 files.

Reproducibility: yes. Current main has a source-clear path where heartbeat returns non-ack pending final text, and the PR includes sanitized live incident context plus after-fix Telegram terminal proof; I did not rerun the live Telegram scenario.

Review metrics: 1 noteworthy metric.

• Heartbeat replay behavior: 1 direct replay path removed; 0 new config knobs. This is an unconditional message-delivery behavior change, so maintainers should consciously accept the recovery tradeoff before merge.

Merge readiness
Overall: 🦞 diamond lobster
Proof: 🦞 diamond lobster
Patch quality: 🦞 diamond lobster
Result: ready for maintainer review.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Mantis proof suggestion
A live Telegram transcript would materially confirm the message-delivery invariant that stale pending finals are not emitted to the heartbeat target. A maintainer can ask Mantis to capture proof by posting a new PR comment that starts with the OpenClaw Mantis account mention, followed by:

telegram live QA: verify a heartbeat with pending stale final text sends only the heartbeat result and does not replay prior topic text.

Risk before merge

• Suppressing heartbeat replay fixes wrong-target leakage, but it can leave a non-ack pending final durable until a non-heartbeat or future route-aware recovery path handles it.
• The inspected PR head still had required checks in progress, so merge should wait for normal CI completion.

Maintainer options:

1. Merge after delivery acceptance (recommended)
   Once required checks pass, maintainers can accept that heartbeats suppress non-ack pending-final replay and rely on non-heartbeat or future route-aware recovery for durable final delivery.
2. Require route-aware recovery first
   If heartbeat recovery must preserve pending finals, require the PR to deliver through the captured pendingFinalDeliveryContext instead of the heartbeat default target.

Next step before merge
No repair lane is needed because no actionable code finding remains; maintainers need to accept the heartbeat delivery tradeoff and wait for required checks.

Security
Cleared: The diff only changes heartbeat/reply runtime logic and tests; I found no dependency, workflow, secret, or supply-chain regression in the patch.

Added the live Telegram reproduction summary to the PR body and linked this to #74257. I kept the reproduction sanitized because the original live trace included private deployment/customer data, but the important shape is captured: one topic gets the intended task reply, then a heartbeat/default route re-emits stale prior task output into another Telegram context in the same minute.

I rechecked this PR against current origin/main on 2026-05-26. The branch still has unique patch content, but GitHub currently reports it as merge-conflicted, so a rebase/refresh is likely needed before maintainer review.

@clawsweeper re-review to refresh the current author-facing status and blockers, please.

🦞👀
ClawSweeper picked this up.

Command router queued. I will update this comment with the next step.

I have refactored this branch against the current origin/main with a merge commit, resolving any PR conflicts without rewriting the branch history.

I have retained both sides of the conflicted heartbeat changes: the current main's pending-final sanitization coverage and this PR's stale final replay suppression. Additionally, I have preserved the newer reply-run registry skip path and wired the isolated heartbeat active-run guard through the current isReplyRunActive path.

I have verified the changes on the head branch fb7eaba67a0ccb08feb6b483a821c9c4ae9abfdd using the following commands:

pnpm test src/auto-reply/reply/get-reply.fast-path.test.ts src/infra/heartbeat-runner.skips-busy-session-lane.test.ts
pnpm exec oxfmt --check --threads=1 src/auto-reply/get-reply-options.types.ts src/auto-reply/reply/get-reply.fast-path.test.ts src/auto-reply/reply/get-reply.ts src/infra/heartbeat-runner.skips-busy-session-lane.test.ts src/infra/heartbeat-runner.test-utils.ts src/infra/heartbeat-runner.ts

Please review the changes.

@clawsweeper

The branch has been narrowed against the current origin/main and the unrelated generated/browser/runtime churn flagged by ClawSweeper has been removed. The PR diff is now limited to the heartbeat replay suppression and its focused tests.

Head now: 78e770c896334a7a75e42dcbf697d6b8c5870df1

Current PR diff files:

• src/auto-reply/get-reply-options.types.ts
• src/auto-reply/reply/get-reply.fast-path.test.ts
• src/auto-reply/reply/get-reply.ts
• src/infra/heartbeat-runner.skips-busy-session-lane.test.ts
• src/infra/heartbeat-runner.test-utils.ts
• src/infra/heartbeat-runner.ts

Validation run locally:

/tmp/openclaw-pr76920-merge/node_modules/.bin/oxfmt --check --threads=1 \\
src/auto-reply/get-reply-options.types.ts \\
src/auto-reply/reply/get-reply.fast-path.test.ts \\
src/auto-reply/reply/get-reply.ts \\
src/infra/heartbeat-runner.skips-busy-session-lane.test.ts \\
src/infra/heartbeat-runner.test-utils.ts \\
src/infra/heartbeat-runner.ts
node scripts/test-projects.mjs \\
src/auto-reply/reply/get-reply.fast-path.test.ts \\
src/infra/heartbeat-runner.skips-busy-session-lane.test.ts

Results: formatter passed; focused Vitest passed 2 shards / 29 tests.

@clawsweeper re-review

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

• State: Failed
• Detail: The targeted re-review did not finish cleanly. Check the workflow run for details.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/26534726490
• Updated: 2026-05-27T19:48:59.534Z

Updated this PR to address the latest ClawSweeper finding.

Head: c2443d2090a7bb46cdebd7e18e6c928b83868c5c7

What changed:

• Restored the active-run compatibility coverage that ClawSweeper called out in src/infra/heartbeat-runner.skips-busy-session-lane.test.ts.
• Re-added the same-agent scheduled heartbeat deferral case.
• Re-added the unscoped active-key ignored case.
• Re-added the immediate heartbeat wake bypass case.
• Runtime code unchanged; this is test coverage only.

Reviewer/bot feedback addressed:

• ClawSweeper P3: “Restore the active-run compatibility cases”.

Proof:

• public artifact or excerpt: this PR diff now includes the restored tests directly in src/infra/heartbeat-runner.skips-busy-session-lane.test.ts.
• raw local audit source, if any: none needed.
• proof limitation, if any: local Vitest and tsgo commands timed out during runner startup in the temporary worktree before producing assertion output; CI is the authoritative execution proof for this pushed head.

Validation:

pnpm exec oxfmt --write src/infra/heartbeat-runner.skips-busy-session-lane.test.ts
pnpm exec oxfmt --check --threads=1 src/infra/heartbeat-runner.skips-busy-session-lane.test.ts
git diff --check

timeout -k5s 90s pnpm vitest src/infra/heartbeat-runner.skips-busy-session-lane.test.ts -t "active reply run|immediate heartbeat wakes|unscoped active reply"
timeout -k5s 120s node scripts/run-vitest.mjs src/infra/heartbeat-runner.skips-busy-session-lane.test.ts -t "active reply run|immediate heartbeat wakes|unscoped active reply"
timeout -k5s 30s node scripts/run-tsgo.mjs -p test/tsconfig/tsconfig.core.test.json --noEmit --pretty false

Result:

• formatter write/check passed
• git diff --check passed
• focused Vitest timed out during local runner startup with Rolldown timing warnings only
• scripts/run-vitest.mjs timed out without test output
• run-tsgo.mjs timed out before output
• CI is running for the pushed head

Current state:

• conflicts: none known before this push; CI will verify current merge state
• CI: running
• proof: supplied / sufficient from prior review; this update is a test-only feedback repair
• rating target: platinum minimum, diamond preferred
• current rating before this push: platinum hermit, but durable verdict still had needs-changes because of the restored-test finding
• remaining blocker, if any: CI/re-review result

@clawsweeper re-review

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Re-review progress:

• State: Complete
• Detail: The targeted re-review finished, the durable review comment was updated, and the synced verdict was routed.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/26534877761
• Updated: 2026-05-27T19:57:13.498Z

Refreshed this branch against current origin/main with a merge commit: b2a620dd5044e975aba215edd2edbbb0437814b1.

This was a CI cleanup refresh, not a behavior expansion. The effective PR diff is still limited to the six heartbeat/reply files, and the stale CI failures are addressed by carrying current main forward:

• the previous Number.NaN oxlint failures are gone
• the previous image-tool.test.ts inboundRoots type error is gone
• GitHub should rerun the failed CI shards on the refreshed head

Validation run locally on the refreshed head:

pnpm exec oxfmt --check --threads=1 src/auto-reply/get-reply-options.types.ts src/auto-reply/reply/get-reply.fast-path.test.ts src/auto-reply/reply/get-reply.ts src/infra/heartbeat-runner.skips-busy-session-lane.test.ts src/infra/heartbeat-runner.test-utils.ts src/infra/heartbeat-runner.ts
pnpm lint --threads=8
pnpm run lint:extensions:bundled
pnpm check:test-types
pnpm test src/auto-reply/reply/get-reply.fast-path.test.ts src/infra/heartbeat-runner.skips-busy-session-lane.test.ts

All passed on b2a620dd5044e975aba215edd2edbbb0437814b1.

@clawsweeper re-review

🦞🧹
ClawSweeper re-review requested.

I asked ClawSweeper to review this item again.
Action: item re-review queued (workflow sweep.yml, event repository_dispatch).
Result: the existing ClawSweeper review comment will be edited in place when the review finishes.

Live Telegram heartbeat stale-final proof, head b2a620dd5044e975aba215edd2edbbb0437814b1

I added fresh staging proof for the heartbeat stale-final case using @kesslerClawBot in the real Telegram staging forum group.

Environment:

OpenClaw PR head: b2a620dd5044e975aba215edd2edbbb0437814b1
Gateway/proof worktree: /tmp/openclaw-pr81313
Staging group: -1003908474243
#General topic: 1
Bot: @kesslerClawBot

Proof shape: seed a temporary session store with a non-ack pending final marker, run runHeartbeatOnce, route heartbeat delivery to the staging Telegram topic, stub the heartbeat model reply to HEARTBEAT_OK, and use the real Telegram sender through the heartbeat outbound adapter.

Observed transcript:

head=b2a620dd5044e975aba215edd2edbbb0437814b1
sessionKey=agent:main:main
chat=-1003908474243
topic=1
staleMarker=PR81313_STALE_FINAL_MUST_NOT_SEND_1779920060790
before.pendingFinalDelivery=true
before.pendingFinalDeliveryText=PR81313_STALE_FINAL_MUST_NOT_SEND_1779920060790
getReplyFromConfig.isHeartbeat=true
getReplyFromConfig.suppressPendingFinalDeliveryReplay=true
[telegram] outbound send ok accountId=default chatId=-1003908474243 messageId=18 operation=sendMessage deliveryKind=text chunkCount=1
heartbeatResult={"status":"ran","durationMs":891}
sent[0].to=-1003908474243
sent[0].text=HEARTBEAT_OK
sent[0].messageThreadId=1
sent[0].messageId=18
sentCount=1
staleMarkerSent=false
after.pendingFinalDelivery=true
after.pendingFinalDeliveryText=PR81313_STALE_FINAL_MUST_NOT_SEND_1779920060790
after.pendingFinalDeliveryAttemptCount=0

What this proves:

• Heartbeat called the reply path with suppressPendingFinalDeliveryReplay=true.
• The only live Telegram send was HEARTBEAT_OK to the staging topic.
• The stale pending-final marker was not sent.
• The pending final remained durable with attempt count 0, so heartbeat did not consume or replay it.

I also refreshed the PR body proof section with this transcript.

@clawsweeper re-review

🦞👀
ClawSweeper picked this up.

Command router queued. I will update this comment with the next step.

Re-review progress:

• State: Superseded
• Detail: A newer re-review for this item started before this run finished, so GitHub cancelled this older run. Check the latest ClawSweeper run for the current result.
• Run: https://github.com/openclaw/clawsweeper/actions/runs/26541968504
• Updated: 2026-05-27T22:20:43.679Z

Landing proof for head e1d6f32:

Behavior addressed: heartbeat runs no longer directly replay non-ack durable pending final text; ack-only heartbeat pending state still clears, and isolated heartbeat runs now skip when their reply run is active.

Real environment tested: local OpenClaw checkout on the PR branch plus GitHub CI for the PR head.

Exact steps or command run after this patch:

• node scripts/run-vitest.mjs src/auto-reply/reply/get-reply.fast-path.test.ts src/infra/heartbeat-runner.skips-busy-session-lane.test.ts
• git diff --check
• /Users/steipete/Projects/agent-scripts/skills/autoreview/scripts/autoreview --mode local
• /Users/steipete/Projects/agent-scripts/skills/autoreview/scripts/autoreview --mode branch --base origin/main

Evidence after fix:

• Local focused Vitest: 2 files, 31 tests passed.
• Local diff whitespace check: clean.
• Autoreview local: clean, no accepted/actionable findings.
• Autoreview branch vs origin/main: clean, no accepted/actionable findings.
• GitHub CI run 26543804437: success on head e1d6f32.
• CodeQL run 26543804438: success.
• CodeQL Critical Quality run 26543804441: success.
• OpenGrep PR Diff run 26543804440 rerun job 78197443511: success.
• Real behavior proof run 26544027357: success.

Observed result after fix: stale non-ack pending final text remains durable but is not returned as the heartbeat reply; heartbeat ack-only pending delivery is still cleared; heartbeat runner passes only normal heartbeat options and avoids replacing an active isolated heartbeat session.

What was not tested: no new live production Telegram incident replay was run during landing; the PR body retains the existing sanitized live/staging proof.