docs(devlog): close out the priority-65 unit

[lidge-jun]

Summary

Closes out the priority-65 unit by recording what each work-phase actually produced. Documentation only — no src/, gui/, or tests/ changes.

Three PRs landed from this unit:

PR	squash	contents
#3471	4968d0f26	responses tool-result boundary (#3259) + combo context-cap failover (carried from #3461)
#3474	00834d710	Claude combo metadata (carried from #3332) + reset-credit identity (#3375 axis D) + two gate-blocked carries (#3327, #3251)
#3477	413227888	rollback journal deletion (part of #3379)

What these documents record

The outcome files keep the parts that would otherwise be lost, which are mostly the places where a plan turned out to be wrong:

• wp3 — the carried PR mapped an OUTPUT ceiling onto an INPUT field, collapsing a 1M combo window to 128k. Verified in three stages, not two: the sniper assertion was confirmed still-red with the defective mapping applied before being fixed, so it demonstrably targets the defect rather than the feature.
• wp4 — the plan said "squash merge fix(combos): fail over provider-specific context caps #3461". Checking immediately before acting showed it was a fork PR whose head had only the four gate checks; Cross-platform CI had never run on it. Merging on gate-green would have violated this unit's own exact-head-CI criterion, so it became a carry.
• wp5 — opening the ledger fails closed and settling fails open, because a reset credit is irreversible and the two directions produce double-spend by opposite routes.
• wp7 — GUI tests were written after the component, so they were proven non-vacuous by reverting the component and confirming exactly four failures. A pre-existing key-set assertion caught the new deletable field; the key list was extended rather than the assertion loosened.
• wp8 — dispositions for the seven items that were examined and deliberately not implemented, each with the evidence that decided it.

Regression proof

	SHA	Cross-platform CI
session start	2421e44ce	failure — macos, tests/codex-auth-context.test.ts:1461
final	413227888	success

dev was already red when this work started, which is why the baseline was captured first — otherwise "did I break it" is unanswerable. All three squash SHAs verified as ancestors of the final origin/dev.

Also recorded: macOS failed twice during this unit and passed on rerun both times, on tests/codex-shim.test.ts and tests/lab-fabric-task.test.ts. Neither file is in any of these diffs, and both pass locally (77/77 and 49/49). Two timing-sensitive tests are intermittently unstable on the macOS runner; that is worth its own look.

Verification

bun run typecheck exit 0, bun run privacy:scan passed. The full local suite was not run at any point in this unit.

Checklist

• Documentation only; no runtime change
• bun x tsc --noEmit clean
• bun run privacy:scan clean
• No GUI change (no screenshot required)

Summary by CodeRabbit

• Bug Fixes

  • Corrected token limit handling so combined input windows no longer collapse unexpectedly when output limits are configured.
  • Preserved accurate tier display behavior, including cases where the response tier is assumed rather than confirmed.
  • Improved rollback journal deletion handling with conflict protection and durable tombstone records.

• Documentation

  • Added closeout records covering completed work packages, verification results, regression checks, and platform test outcomes.
  • Documented ledger failure behavior for opening and closing operations across retry scenarios.

---

The carries this unit landed are attributed here as well, since this description names them:

Co-authored-by: RHODIZ IT info.rhodiz@gmail.com
Co-authored-by: full999 daiki.furutani@walker-s.co.jp
Co-authored-by: olddonkey olddonkeyblog@gmail.com
Co-authored-by: Abhishek Sharma abhicse24@gmail.com

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-04T16:16:26.799377Z	4607d74	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[github-actions]

⚠️ Deterministic hygiene checks failed.

• missing_coauthor_credit — This pull request says it reimplements, supersedes, carries, or rebases another author's pull request, but no Co-authored-by trailer names that author. Prose in a commit body is not read by anything; the trailer is what GitHub counts. Add it to the description or a commit, or obtain attribution-approved. Paths: #3251, #3327, #3332, #3461.

[github-actions]

✅ READY

• all PR quality gates passed.

Hygiene

✅ Deterministic PR hygiene checks passed.

[coderabbitai]

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Path: .coderabbit.yaml

Review profile: ASSERTIVE

Plan: Team

Run ID: 5c6f2fb1-5f75-4bd5-b4d8-9304e14a5779

📥 Commits

Reviewing files that changed from the base of the PR and between 9c0e3ca and 4607d74.

📒 Files selected for processing (6)

• devlog/_plan/260904_priority65_closeout/080_outcome.md
• devlog/_plan/260904_priority65_closeout/081_outcome_wp4.md
• devlog/_plan/260904_priority65_closeout/082_outcome_wp5.md
• devlog/_plan/260904_priority65_closeout/083_outcome_wp6.md
• devlog/_plan/260904_priority65_closeout/084_outcome_wp7.md
• devlog/_plan/260904_priority65_closeout/085_outcome_wp8.md

Included review availability: Your plan provides up to 10 included reviews per hour; 3 remain after this review.

---

📝 Walkthrough

Walkthrough

The PR adds six priority-65 closeout documents. They record landed commits, a token-mapping correction, gate verification, ledger behavior, rollback journal validation, issue dispositions, CI results, and observed macOS test flakes.

Changes

Priority 65 closeout

Layer / File(s)	Summary
Landing records and wp3 correction devlog/_plan/260904_priority65_closeout/080_outcome.md	Records three landed PRs and ancestor checks. Documents the correction from maxInputTokens to maxOutputTokens and the three-stage verification.
Work package 4 through 6 outcomes devlog/_plan/260904_priority65_closeout/081_outcome_wp4.md, devlog/_plan/260904_priority65_closeout/082_outcome_wp5.md, devlog/_plan/260904_priority65_closeout/083_outcome_wp6.md	Records the wp4 carry and 23/23 matrix result, wp5 ledger failure modes and retry identity behavior, and wp6 gate and screenshot verification.
Work package 7 and 8 closeout devlog/_plan/260904_priority65_closeout/084_outcome_wp7.md, devlog/_plan/260904_priority65_closeout/085_outcome_wp8.md	Records tombstone-based journal deletion, route and GUI validation, issue dispositions, CI comparisons, ancestor checks, macOS flakes, and unrun local suites.

Estimated code review effort: 2 (Simple) | ~10 minutes

Merge Risk: ⚪ Minimal · up to 4607d

This change records Priority 65 outcomes and verification history without changing runtime behavior. No current merge-blocking risk remains.

🚥 Pre-merge checks | ✅ 5

✅ Passed checks (5 passed)

Check name	Status	Explanation
Docstring Coverage	✅ Passed	No functions found in the changed files to evaluate docstring coverage. Skipping docstring coverage check. Docstring coverage is scoped to functions touched by this diff. Analyzed 0 functions across 0…
Linked Issues check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Out of Scope Changes check	✅ Passed	Check skipped because no linked issues were found for this pull request.
Description Check	✅ Passed	Check skipped - CodeRabbit’s high-level summary is enabled.
Title check	✅ Passed	The title accurately identifies the documentation-only devlog closeout for the priority-65 unit. It is concise, specific, and aligned with the documented wp3–wp8 outcome changes.

✨ Finishing Touches 💡 1

🛠️ Fix failing CI checks 💡

• Create stacked PR
• Commit on current branch

🧪 Generate unit tests (beta)

• Create PR with unit tests
• Commit unit tests in branch codex/priority65-outcome

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4607d74740

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[chatgpt-codex-connector]

Move the closed unit out of _plan

This outcome marks the unit as completed and records its final regression result, but every file remains under devlog/_plan/260904_priority65_closeout. That leaves a finished unit advertised as active to maintainers or tooling enumerating _plan, while its historical record is absent from _fin; move the entire unit directory to devlog/_fin/ as part of this closeout.

AGENTS.md reference: AGENTS.md:L83-L86

Useful? React with 👍 / 👎.

[lidge-jun]

리뷰 · 우선순위 40 / 80

이 PR은 코드가 아니라 마무리 기록입니다. 우선순위 65 유닛이 이미 dev에 올린 세 스쿼시(#3471 4968d0f26, #3474 00834d710, #3477 413227888)가 무엇을 했는지, 계획이 어디서 틀렸는지를 devlog/_plan/260904_priority65_closeout/080~085 여섯 파일에만 남깁니다. src/, gui/, tests/는 건드리지 않습니다.

바로 앞 PR #3485는 같은 제목으로 열렸다가 닫혔습니다. 그 브랜치에는 이미 dev에 스쿼시된 커밋이 남아 있어서 diff가 GUI 18개 파일까지 끌고 와 스크린샷 게이트가 걸렸고, 작성자가 현재 dev에서 새로 자른 codex/priority65-outcome으로 이 #3486만 다시 열었습니다. 지금 HEAD(9c0e3ca80)에는 계획 문서 000~070이 이미 #3471과 함께 들어와 있고, 이 PR은 그 위에 결과 문서만 얹습니다. 세 스쿼시 SHA는 전부 현재 origin/dev의 조상입니다.

문서가 말하는 내용과 현재 코드도 맞습니다. wp2 responses 경계는 src/server/responses/core.ts에 번역 경로 가드가 있고, wp3 Claude combo 메타는 src/adapters/anthropic.ts / src/codex/catalog/provider-fetch.ts / src/providers/registry.ts 쪽, wp4 context-cap failover는 src/combos/failover.ts, wp5 reset-credit operationId는 src/codex/auth-api.ts와 CLI src/cli/account-auth.ts, wp7 저널 삭제는 src/integrations/journal.ts의 툼스톤 append와 src/server/management/integration-routes.ts DELETE, GUI RollbackHistory.tsx의 deletable 필드가 HEAD에 있습니다. 080이 강조하는 wp3 실수(OUTPUT maxTokens를 INPUT maxInputTokens에 매핑해 1M 창이 128k로 무너진 일)와 wp4에서 fork PR 게이트 4개만 초록인 채 머지하지 않고 carry로 바꾼 판단은 나중에 같은 함정을 피하려고 남기는 값입니다.

다만 게이트가 한 가지 막혀 있습니다. 라벨 intake: hygiene-blocked와 hygiene 봇이 missing_coauthor_credit을 걸었습니다. 본문이 #3251 / #3327 / #3332 / #3461을 carry했다고 읽히는데, 이 PR 커밋에는 해당 기여자의 Co-authored-by 트레일러가 없습니다. 실제 코드 carry 크레딧은 이미 #3471·#3474 스쿼시에 들어갔고, 여기는 그 일을 되짚는 문서뿐입니다. 그래도 봇은 본문 문구만 보고 막습니다. PR은 지금 draft입니다.

라인별 문제는 런타임 버그가 아니라 머지 전 정리입니다.

080_outcome.md - 랜딩 표와 조상 확인 서술은 HEAD와 맞고, 문서만으로는 런타임 회귀를 만들지 않는다
081_outcome_wp4.md - fork PR #3461을 게이트만으로 머지하지 않은 이유를 남긴 것은 현재 exact-head-CI 방향과 같다
082_outcome_wp5.md - 열기 fail-closed / 닫기 fail-open 서술은 auth-api.ts의 현재 계약과 맞다
083_outcome_wp6.md / 084_outcome_wp7.md - 게이트 해제·툼스톤·비공허 테스트 증명은 이미 랜딩된 #3474/#3477을 정확히 되짚는다
085_outcome_wp8.md - 처분 표와 macOS flake 관찰은 기록 가치는 있으나, flake 두 건(codex-shim, lab-fabric-task)은 이 PR 범위 밖이다
PR 본문 / hygiene - carried from 문구가 missing_coauthor_credit을 다시 켠다. 코드 carry가 아닌 결과 기록이면 본문을 그 뜻으로 고치거나, 이미 랜딩된 크레딧을 인정하는 attribution-approved를 쓰거나, 해당 기여자 트레일러를 커밋에 넣어야 draft가 풀린다
#3485 - 같은 주제의 이전 PR은 이미 superseded로 닫혔고, 이 #3486이 대체본이다. #3485에 리뷰를 덧붙일 필요는 없다

메인테이너의 판단이 필요한 지점

• hygiene를 본문 수정으로 풀지, attribution-approved로 풀지, Co-authored-by를 문서 커밋에까지 다시 넣을지
• 085에 적힌 macOS flake 두 건을 별도 이슈로 바로 뺄지, 다음 유닛으로 미룰지
• 계획 문서 000~070은 이미 HEAD에 있는데, 결과 문서만 따로 머지하는 이 분리가 앞으로도 기본 패턴인지

너의 추천
hygiene만 풀면 머지해도 됩니다. 가장 싼 길은 PR 본문에서 carried from #3251/#3327/#3332/#3461처럼 이 PR이 그 코드를 다시 싣는 것처럼 읽히는 문장을 이미 #3471/#3474에 랜딩된 carry를 기록한다로 고치는 것입니다. 그게 싫으면 attribution-approved를 붙이세요. 닫을 이유는 없고, #3485는 그대로 닫힌 채 두면 됩니다.

이 댓글은 grok-bot이 작성했습니다

[Ingwannu]

Reviewed exact head 4607d74. The replacement branch is now based directly on current dev and contains only the six intended priority-65 outcome records. The recorded landed PR/SHA relationships and implementation summaries match the current dev ancestry, the prior 64-file replay problem is gone, deterministic hygiene is green after attribution was added, and CodeRabbit reports no actionable findings. Documentation-only and safe to merge under the normal branch policy.