[juliusmarminge]

A PR watch holds an edit of a remark the agent already saw until a read finds it unchanged, then wakes once with the final text. The review finding was correct: the once-per-head rule (remarkEditsTold) dropped every later edit of a remark already reported on that head, so a review bot's 'reviewing' -> 'finished' after a push delivered only 'reviewing'. Instead of removing remarkEditsTold (which would leave nothing beyond #30674's settle delay), an edit of a remark already reported on this head is now held like any other edit, and delivered once its editedAt is TOLD_EDIT_SETTLE_MS (10 minutes) old. That is longer than a bot's countdown rewrites, so a countdown stays quiet until it stops and then wakes once with its final text, while a finished review on the same head still reaches the agent. A remark's first edit on a head still settles after one unchanged read, and the record still resets after a push. Tradeoffs: the final text of an already-reported remark arrives about 10 minutes after the bot's last edit, and while such an edit is held the reactor re-reads each 2-minute sweep even under a quiet host fingerprint. The per-head regression test is rewritten around timed reads to cover the countdown, a second remark, the reviewer's reviewing -> finished sequence after a push, and the earlier old-head held-edit case; it fails on the previous code.

Fixes #30342

Supersedes #36280. The change is @royalaid's; their commits are kept.

Reproduction on main

1. In a T3 Code thread, ask the agent to watch an open GitHub PR using watch_pull_request. Keep checks and the PR head unchanged. 2. Have another account post a comment and allow the watch to deliver it. 3. Edit that same comment, then leave it unchanged for at least two watch sweeps, approximately four minutes. 4. Repeat the edit-and-wait sequence several times without pushing. Main generates another new-comment wake for each observed edit; this change delivers the first settled edit and suppresses subsequent edits of that comment on that head. 5. Push a new commit, edit the same comment again, and leave it unchanged for two sweeps: another edited-comment wake should arrive. These steps were derived from the diff and issue, not executed.

Evidence

Being captured: before and after in the real app. This section is filled in before the pull request can merge.

Tests

vp test run apps/server/src/orchestration-v2/pullRequestWatch.test.ts: the new test fails on f6f3323's evaluator (final edit never reported); 42/42 pass with the fix. vp run check:changed: format, lint, knip, graphql, typecheck (37 packages) and tests (8 files) all pass, after formatting the test and making the new constant module-private because knip flagged its export as unused.

Not in this change

• runtimeLayer.test.ts 'wakes a watched thread once' cases depend on test order in the shared layer and fail on main too when filtered with -t "watch"; not caused by this PR.
• fix(server): a PR watch wakes once for a comment's edits, not on every edit #30674 is still open; this branch contains its commits, so landing this one supersedes it.
• fix(server): a PR watch wakes once for a comment's edits, not on every edit #30674 is still open and this branch carries its commits; landing this PR supersedes it.
• The longer settle is measured from the host's editedAt against the server clock; large clock skew between host and server would shorten or lengthen it.

Who did the work

• Written by Claude Opus 5.5, in Claude Code
• Reviewed by GPT-6.1 Sol, in Codex, and by Claude Opus 5.5, each on its own
• Proof captured by GPT-6.1 Sol and checked by Claude Opus 5.5

Run by the t3code factory in T3 Code.

[github-actions]

Thread transfer impact

✅ Thread transfer remains within every enforced ceiling.

Provider	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Codex	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Codex	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	21.5 KiB	21.5 KiB	0 B (0.0%)	24.4 KiB	✅
Codex	Live turn messages	1	1	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	0 B (0.0%)	6.8 KiB	✅
Claude	Thread snapshot wire	3.8 KiB	3.8 KiB	0 B (0.0%)	4.9 KiB	✅
Claude	Live turn WebSocket wire	1.2 KiB	1.2 KiB	0 B (0.0%)	2.0 KiB	✅
Claude	Live turn WebSocket decoded	21.9 KiB	21.9 KiB	0 B (0.0%)	24.4 KiB	✅
Claude	Live turn messages	1	1	0 (0.0%)	8	✅

Other scenarios

All 64 scenario measurements match main.

All 64 scenario measurements

Provider	Scenario	Metric	Main baseline	This PR	Impact	PR ceiling
Codex	short turn resume	WebSocket decoded	10.5 KiB	10.5 KiB	0 B (0.0%)	12.7 KiB	✅
Codex	short turn resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Codex	long turn running, cold open	HTTP decoded	281.7 KiB	281.7 KiB	0 B (0.0%)	302.7 KiB	✅
Codex	long turn running, cold open	rows	75	75	0 (0.0%)	75	✅
Codex	long turn running, history page	HTTP decoded	92.4 KiB	92.4 KiB	0 B (0.0%)	107.4 KiB	✅
Codex	long turn running, history page	rows	75	75	0 (0.0%)	75	✅
Codex	long turn finished, cold open	HTTP decoded	381.5 KiB	381.5 KiB	0 B (0.0%)	410.2 KiB	✅
Codex	long turn finished, cold open	rows	151	151	0 (0.0%)	151	✅
Codex	long turn finished, socket cold open	WebSocket decoded	381.6 KiB	381.6 KiB	0 B (0.0%)	410.2 KiB	✅
Codex	long turn finished, socket cold open	WebSocket messages	2	2	0 (0.0%)	2	✅
Codex	long turn finished, socket cold open	rows	151	151	0 (0.0%)	151	✅
Codex	long turn finished, work page	HTTP decoded	186.2 KiB	186.2 KiB	0 B (0.0%)	200.2 KiB	✅
Codex	long turn finished, work page	rows	150	150	0 (0.0%)	150	✅
Codex	idle thread resume	WebSocket decoded	65 B	65 B	0 B (0.0%)	200 B	✅
Codex	idle thread resume	WebSocket messages	1	1	0 (0.0%)	1	✅
Codex	fleet shell snapshot	HTTP decoded	42.4 KiB	42.4 KiB	0 B (0.0%)	53.7 KiB	✅
Codex	fleet shell snapshot	rows	27	27	0 (0.0%)	27	✅
Codex	fleet shell subscribe	WebSocket decoded	65 B	65 B	0 B (0.0%)	200 B	✅
Codex	fleet shell subscribe	WebSocket messages	1	1	0 (0.0%)	1	✅
Codex	fleet shell resume check	HTTP decoded	18 B	18 B	0 B (0.0%)	100 B	✅
Codex	fleet shell resume check	rows	0	0	0	0	✅
Codex	stale shell reload	HTTP decoded	42.4 KiB	42.4 KiB	0 B (0.0%)	53.7 KiB	✅
Codex	stale shell reload	rows	27	27	0 (0.0%)	27	✅
Codex	stale shell resume	WebSocket decoded	140 B	140 B	0 B (0.0%)	400 B	✅
Codex	stale shell resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Codex	stale todo list cold open	HTTP decoded	72.4 KiB	72.4 KiB	0 B (0.0%)	83.0 KiB	✅
Codex	stale todo list cold open	rows	75	75	0 (0.0%)	75	✅
Codex	stale todo list cold open	nodes	2	2	0 (0.0%)	4	✅
Codex	stale todo list cold open	subagents	2	2	0 (0.0%)	4	✅
Codex	stale todo list cold open	plans	1	1	0 (0.0%)	2	✅
Codex	stale todo list cold open	providerTurns	1	1	0 (0.0%)	2	✅
Codex	stale todo list cold open	checkpoints	1	1	0 (0.0%)	2	✅
Claude	short turn resume	WebSocket decoded	10.6 KiB	10.6 KiB	0 B (0.0%)	12.7 KiB	✅
Claude	short turn resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Claude	long turn running, cold open	HTTP decoded	283.3 KiB	283.3 KiB	0 B (0.0%)	302.7 KiB	✅
Claude	long turn running, cold open	rows	75	75	0 (0.0%)	75	✅
Claude	long turn running, history page	HTTP decoded	93.4 KiB	93.4 KiB	0 B (0.0%)	107.4 KiB	✅
Claude	long turn running, history page	rows	75	75	0 (0.0%)	75	✅
Claude	long turn finished, cold open	HTTP decoded	384.2 KiB	384.2 KiB	0 B (0.0%)	410.2 KiB	✅
Claude	long turn finished, cold open	rows	151	151	0 (0.0%)	151	✅
Claude	long turn finished, socket cold open	WebSocket decoded	384.3 KiB	384.3 KiB	0 B (0.0%)	410.2 KiB	✅
Claude	long turn finished, socket cold open	WebSocket messages	2	2	0 (0.0%)	2	✅
Claude	long turn finished, socket cold open	rows	151	151	0 (0.0%)	151	✅
Claude	long turn finished, work page	HTTP decoded	188.1 KiB	188.1 KiB	0 B (0.0%)	200.2 KiB	✅
Claude	long turn finished, work page	rows	150	150	0 (0.0%)	150	✅
Claude	idle thread resume	WebSocket decoded	66 B	66 B	0 B (0.0%)	200 B	✅
Claude	idle thread resume	WebSocket messages	1	1	0 (0.0%)	1	✅
Claude	fleet shell snapshot	HTTP decoded	43.0 KiB	43.0 KiB	0 B (0.0%)	53.7 KiB	✅
Claude	fleet shell snapshot	rows	27	27	0 (0.0%)	27	✅
Claude	fleet shell subscribe	WebSocket decoded	66 B	66 B	0 B (0.0%)	200 B	✅
Claude	fleet shell subscribe	WebSocket messages	1	1	0 (0.0%)	1	✅
Claude	fleet shell resume check	HTTP decoded	18 B	18 B	0 B (0.0%)	100 B	✅
Claude	fleet shell resume check	rows	0	0	0	0	✅
Claude	stale shell reload	HTTP decoded	43.0 KiB	43.0 KiB	0 B (0.0%)	53.7 KiB	✅
Claude	stale shell reload	rows	27	27	0 (0.0%)	27	✅
Claude	stale shell resume	WebSocket decoded	140 B	140 B	0 B (0.0%)	400 B	✅
Claude	stale shell resume	WebSocket messages	2	2	0 (0.0%)	2	✅
Claude	stale todo list cold open	HTTP decoded	73.4 KiB	73.4 KiB	0 B (0.0%)	83.0 KiB	✅
Claude	stale todo list cold open	rows	75	75	0 (0.0%)	75	✅
Claude	stale todo list cold open	nodes	2	2	0 (0.0%)	4	✅
Claude	stale todo list cold open	subagents	2	2	0 (0.0%)	4	✅
Claude	stale todo list cold open	plans	1	1	0 (0.0%)	2	✅
Claude	stale todo list cold open	providerTurns	1	1	0 (0.0%)	2	✅
Claude	stale todo list cold open	checkpoints	1	1	0 (0.0%)	2	✅

Baseline: e94db17 · PR result: 0b01bd0 · Source CI: success

Scenario and decoded snapshot size

10 historical turns, 5 command tools per turn, 878.9 KiB retained MCP result per historical turn, and a 1.05 MiB retained result in the measured turn.

• Codex decoded thread snapshot: 108.5 KiB
• Claude decoded thread snapshot: 108.8 KiB

Updated in place by a trusted workflow. PR artifacts are strictly validated and never executed.

[macroscopeapp]

Approvability

Verdict: Approved at 0b01bd0

Macroscope's review found this PR approvable — This change batches repeated PR-comment edits before notifying the agent and preserves pending edits across watch restarts. Its effects are limited to notification timing and host reads, with no new host-side writes or access changes.

^{You can add or adjust custom eligibility rules. Learn more.}

[coderabbitai]

No actionable comments were generated in the recent review.

Walkthrough

Pull-request watches now distinguish new remarks from edits. They retain changing edits across reads and watch restarts, then report confirmed edits separately. GitCafe activity now supplies edit timestamps when a comment’s update time is later than its creation time.

Changes: Pull-request remark edit tracking

Layer / File(s)	Summary
Edit timestamps and watch state contract packages/source-control-gitcafe/src/server/gitCafePullRequestJson.ts, packages/source-control-gitcafe/src/server/gitCafePullRequestJson.test.ts, packages/contracts/src/threadPullRequest.ts	GitCafe remarks now include editedAt when updatedAt is later than createdAt. The watch state can store pending edits and IDs of remarks whose edits were reported.
Classify and report settled edits apps/server/src/orchestration-v2/pullRequestWatch.ts, apps/server/src/orchestration-v2/pullRequestWatch.test.ts	The evaluator waits for an edit timestamp to remain unchanged before reporting an edit. It labels edits separately from new remarks, suppresses repeated edits to a reported remark during the 10-minute settle interval, and clears held edit state when the PR head changes. Tests cover settling, new remarks, and head changes.
Read, persist, and resume pending edits apps/server/src/orchestration-v2/PullRequestWatchReactor.ts, apps/server/src/orchestration-v2/Orchestrator.ts, apps/server/src/orchestration-v2/runtimeLayer.test.ts	Pending edits trigger reads and cache invalidation. Watch equality includes edit state, and a restarted watch copies that state forward. Runtime tests cover quiet fingerprints, restarts, and paginated review threads.

Estimated code review effort: 3 (Moderate) | ~25 minutes

Change: Bug fix · Severity of issue fixed: Medium

Sequence Diagram(s)

sequenceDiagram
  participant Provider
  participant GitCafePullRequestJson
  participant PullRequestWatchReactor
  participant evaluatePullRequestWatch
  Provider->>GitCafePullRequestJson: Supply comment timestamps
  GitCafePullRequestJson->>PullRequestWatchReactor: Return remarks with editedAt
  PullRequestWatchReactor->>evaluatePullRequestWatch: Evaluate remarks and watch state
  evaluatePullRequestWatch-->>PullRequestWatchReactor: Return pending edit state or edited remark change

Loading

Suggested reviewers: shivamhwp, t3dotgg, maria-rcks

Merge Risk: Minimal · up to 0b01b

This change holds edits to already-reported PR comments until they settle, so bot countdown rewrites no longer wake the agent repeatedly. No concrete merge-blocking defect was found, and the one open concern about stale comment reads has no supporting evidence.

Pre-merge checks: 4

Check name	Status	Explanation
Title check		The title says a PR watch reports a comment's edit once per head, which is the bug the change fixes. It names the server component and the main behavior. The PR objectives describe the same change: ed…
Description check		The description follows the template. It has a problem statement, a change section that explains the approach and tradeoffs, a scope section that links issue #30342, and a verification section with fo…
Linked Issues check		[#30342] The issue requires that a PR watch not report an already-delivered comment again as a new comment on each edit. It allows either skipping edits or labelling and coalescing them, for example b…
Out of Scope Changes check		All listed changes support edit detection and settling for PR watches. Orchestrator.ts, PullRequestWatchReactor.ts, pullRequestWatch.ts, and their tests implement the held-edit behavior. threadPullReq…

Full details: Title check

Explanation

The title says a PR watch reports a comment's edit once per head, which is the bug the change fixes. It names the server component and the main behavior. The PR objectives describe the same change: edits already reported on the current head are now held and delivered after a settle period, not reported once per head. The title's wording is a little misleading, because the fix is the opposite of the literal wording: it lets later edits through instead of limiting them to once per head. Still, it refers to the right area and the right behavior, so it is a pass.

Full details: Description check

Explanation

The description follows the template. It has a problem statement, a change section that explains the approach and tradeoffs, a scope section that links issue #30342, and a verification section with focused test results. It also names the agent and harness that did the work. Two items are weak: the Scope section does not say the change is a bug fix for a bug, and the Evidence section is still pending, with a note that it must be filled before merge. The description also repeats the #30674 supersession note twice. These gaps are minor, so the description is mostly complete.

Full details: Linked Issues check

Explanation

[#30342] The issue requires that a PR watch not report an already-delivered comment again as a new comment on each edit. It allows either skipping edits or labelling and coalescing them, for example by delivering only the final state. The change in pullRequestWatch.ts meets the coalescing option. A first edit on a head settles after one unchanged read. Later edits to an already-reported remark are held until editedAt is 10 minutes old, then delivered once with the final text. Changes are tagged with edited: true on PullRequestWatchChange, and edited remarks use a distinct summary. The remarkEditsTold and pendingEdits fields in packages/contracts/src/threadPullRequest.ts store this state. PullRequestWatchReactor.ts forces a read while edits are pending, so held edits are delivered. Orchestrator.ts carries the state across unwatch and rewatch. The test changes in pullRequestWatch.test.ts and runtimeLayer.test.ts cover the countdown case, repeated edits, and a finished review after a push. The reproduction steps in the PR description were not executed. Real-app evidence is still pending. This is not a coding requirement, so it does not change this result. Evidence is taken from the PR summary and the stated test results. No repository inspection was made.

Full details: Out of Scope Changes check

Explanation

All listed changes support edit detection and settling for PR watches. Orchestrator.ts, PullRequestWatchReactor.ts, pullRequestWatch.ts, and their tests implement the held-edit behavior. threadPullRequest.ts adds the persisted fields that this behavior needs. runtimeLayer.test.ts covers the same behavior end to end. The gitCafePullRequestJson.ts and test changes derive editedAt from updatedAt only when it is later than createdAt. This supplies the edit timestamp that the watch uses, so it is part of the same feature. No unrelated change was found in the summary. No repository inspection was made, so the result rests on the summary.

Finishing Touches

Generate unit tests (beta)

• Commit to this branch
• Create a new PR

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Usage-based review receipt

• Mode: Continue automatically
• Reviewed files: 8
• Charged: $2.00
• View usage details

Note

This review was completed with usage-based billing: files reviewed beyond your plan's included limits are billed at $0.25/file. View usage-based billing.

Recent review info

Run configuration

• Source mode: Auto
• Repository: pingdotgg/t3code
• Branch: factory/pr-36280
• File: .coderabbit.config.ts
• Commit: 0b01bd0
• Layers: Repository TS config
• Review profile: CHILL
• Plan: Team
• Run ID: 896a6541-2533-4e8d-94a6-ddfdeb2e73cf

Commits

Reviewing files that changed from the base of the PR and between e94db17 and 0b01bd0.

Files selected for processing: 8

• apps/server/src/orchestration-v2/Orchestrator.ts
• apps/server/src/orchestration-v2/PullRequestWatchReactor.ts
• apps/server/src/orchestration-v2/pullRequestWatch.test.ts
• apps/server/src/orchestration-v2/pullRequestWatch.ts
• apps/server/src/orchestration-v2/runtimeLayer.test.ts
• packages/contracts/src/threadPullRequest.ts
• packages/source-control-gitcafe/src/server/gitCafePullRequestJson.test.ts
• packages/source-control-gitcafe/src/server/gitCafePullRequestJson.ts

Limit details: You’ve used all 10 included reviews currently available.

[juliusmarminge]

The automated pipeline stopped on this pull request and will not merge it:

the fix is written and reviewed, but proving it needs something no worker has:

• Permanent blocker: the live reproduction requires a second authenticated GitHub account to post and repeatedly edit a comment while the first account watches; only one authenticated account was available, and this job prohibits posting forge comments.
• All runtime expectations remain unverified: repeated-edit suppression, notification after a push, and delivery of new comments. No captures were produced or uploaded; ghdrop was not attempted.
• Source inspection shows the supplied head permits subsequent edits on the same head to notify again after a ten-minute settle, which differs from the stated once-per-head expectation. This was not verified at runtime.

It stays open for a person to finish or close.