perf(server): restarts no longer replay PR discovery for every old thread

[t3dotgg]

Every server restart ran pull request discovery again for every settled branch thread with no saved PR. A "no PR" result is not saved, so each boot repeated the full pass: several git spawns and a gh call per thread, and up to 5 lookups per thread when gh fails. On a 5,000-thread scale bench with gh missing, that was about 3,850 git spawns/min for about 5 minutes after each boot.

Fix

The startup backfill from #10101 now only covers threads settled in the last 7 days (BACKFILL_WINDOW).

The check uses settledAt, not updatedAt. Auto-settle sets settledAt to the last activity, but sets updatedAt to the settle time. With updatedAt, one mass auto-settle would put every old thread back in the window for a week of restarts.

Decision: 7-day window or delete the backfill

The backfill was a one-time migration for threads from before #10101 (merged 2026-09-06, shipped in v0.0.39). Since v0.0.39, the 1-minute pass checks every unsettled branch thread, so each thread is checked until it settles. The backfill only matters for a settled thread with no saved PR.

Who	Today	7-day window (this PR)	Delete
Users upgrading from 0.0.38 or older	All settled threads get a lookup	Only threads settled in the last 7 days	No settled thread gets a lookup
Users who fix gh later, or open a PR outside T3 Code after the thread settled	All settled threads found on next restart	Only threads settled in the last 7 days	Not found
Cost on every restart	All settled threads with no PR	Threads settled in the last 7 days with no PR	None

In both lost cases, unsettling the thread runs discovery again. Deleting also removes the retry state in the reactor (about 50 lines) and 2 of its tests.

I kept the window because it is the smaller behavior change. Deleting it is an easy follow-up if we prefer that.

Verification

• vp test run apps/server/src/orchestration/ThreadPullRequestReactor.test.ts: 16 passed, including the unsettled-vs-full-read parity test from #13765.
• The new test seeds a recent settled thread, an old one, and an auto-settled one (old settledAt, recent updatedAt). Only the recent one gets a sync. Keying the check on updatedAt makes it fail.
• vp lint and vp fmt on the changed files, and vp run --filter t3 typecheck.
• I did not re-run the scale bench on this build.

Made by Claude Opus 5.5 (1M context) in Claude Code, running in T3 Code.

🤖 Generated with Claude Code

[macroscopeapp]

Approvability

Verdict: Not approved

Macroscope's review found this PR not approvable — The PR changes production startup behavior by suppressing pull-request discovery for older settled threads, rather than merely optimizing an unchanged path. Although the implementation is small and tested, the cutoff can affect whether significant Git/GitHub processing and PR links are created after restarts.

No code changes detected at 80bd4c8. Prior analysis still applies.

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

[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	13.5 KiB	13.5 KiB	+15 B (+0.1%)	15.1 KiB	✅
Codex	Thread snapshot wire	7.1 KiB	7.1 KiB	+4 B (+0.1%)	7.3 KiB	✅
Codex	Live turn WebSocket wire	6.4 KiB	6.5 KiB	+11 B (+0.2%)	7.8 KiB	✅
Codex	Live turn WebSocket decoded	56.2 KiB	56.3 KiB	+44 B (+0.1%)	66.4 KiB	✅
Codex	Live turn messages	9	10	+1 (+11.1%)	21	✅
Claude	Total thread wire	13.5 KiB	13.5 KiB	+4 B (+0.0%)	15.1 KiB	✅
Claude	Thread snapshot wire	7.1 KiB	7.1 KiB	−9 B (−0.1%)	7.3 KiB	✅
Claude	Live turn WebSocket wire	6.4 KiB	6.5 KiB	+13 B (+0.2%)	7.8 KiB	✅
Claude	Live turn WebSocket decoded	57.0 KiB	57.0 KiB	0 B (0.0%)	66.4 KiB	✅
Claude	Live turn messages	9	9	0 (0.0%)	21	✅

Baseline: 8872666 · PR result: d1273a3 · 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: 114.0 KiB
• Claude decoded thread snapshot: 114.7 KiB

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

[coderabbitai]

Navigate logical layers of code changes, visualize relationships, and explore their blast radius.

No actionable comments were generated in the recent review. 🎉

ℹ️ Recent review info

⚙️ Run configuration

Configuration used: Repository: pingdotgg/t3code/.coderabbit.yaml

Review profile: CHILL

Plan: Team

Run ID: c330807d-ce27-4e63-9ffc-46f8a962bcec

📥 Commits

Reviewing files that changed from the base of the PR and between 80bd4c8 and d1273a3.

📒 Files selected for processing (2)

• apps/server/src/orchestration/ThreadPullRequestReactor.test.ts
• apps/server/src/orchestration/ThreadPullRequestReactor.ts

Included review availability: This review used your included allowance. Your plan provides up to 10 included reviews per hour; 6 remain after this review.

---

📝 Walkthrough

Walkthrough

Startup backfill now queues settled threads without a branch pull request only when their settlement timestamp is within the previous seven days. If settledAt is absent, the filter uses updatedAt. A test checks the cutoff and timestamp behavior.

Changes

Startup backfill

Layer / File(s)	Summary
Filter startup backfill by settlement time apps/server/src/orchestration/ThreadPullRequestReactor.ts, apps/server/src/orchestration/ThreadPullRequestReactor.test.ts	The reactor applies a seven-day cutoff using settledAt, or updatedAt when settledAt is absent. The test checks which threads are synchronized.

Priority: ➖ Normal

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

Change: Bug fix

Suggested reviewers: juliusmarminge

Merge Risk: ⚪ Minimal · up to d1273

The seven-day startup backfill limit has no identified issue requiring a fix before merge.

🚥 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 2…
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.
Title check	✅ Passed	The title clearly and concisely describes the main change: preventing repeated pull-request discovery for old threads after restarts.
Description check	✅ Passed	The description explains the problem, the seven-day settledAt-based fix, the rationale, trade-offs, and verification results. It does not use the template headings or include the checklist, but it is …

✨ Finishing Touches

📝 Generate docstrings

• Commit to this branch
• Create a new PR

🧪 Generate unit tests (beta)

• Commit to this branch
• Create a new PR

---

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