[incognitojam]

Intake batch 30 ended at pingdotgg#22052, which broke Stop for running subagents. Upstream fixed it four days later in pingdotgg#23886, 719 commits further down the queue, so the fork carried the bug until fork #1116 ported the fix by hand. Nothing in the survey pointed at it.

intake-run.ts survey now also lists later upstream fix and revert commits, up to fetched upstream/main, that name a surveyed commit. A later commit counts when its PR title (or subject, for a direct commit) is a fix or revert, and its message or PR description names a surveyed commit by #N, pingdotgg/t3code#N, a pingdotgg/t3code/pull/N link, or "reverts commit ". The text output shows each flagged fix under the commit it names and quotes the sentence that names it. In --json, each survey entry has a laterFixes array. Fixes the surveyed base already imported, going by Upstream-PR/Upstream-Commit trailers, are left out. When the fork tracks a flagged fix in .github/upstream-tracked-prs.json, the output shows its reason. PR descriptions are read with gh api graphql, 100 per query.

Precision matters more than recall here, so the check needs an explicit reference. I tried a second signal, a fix whose title says "again", "restore", or "no longer" and that touches a file a batch commit changed. On batch 30 it matched about 45 later fixes, and none of them fixed a batch commit. SHAs other than in "reverts commit" are ignored because takeover PRs often cite the main SHA they tested against. Links and qualified references to other repositories are ignored too.

The runbook gets one paragraph: check each flagged fix before the batch lands, and when it fixes a regression the batch introduces, extend the batch, import it early, or port it by hand and track the upstream PR.

Validation

I ran survey --base 9a69b6c2ba --through fb990765a4 for the real batch-30 range: 55 commits, with 1,424 later upstream commits checked. It flags 9 later fixes:

Batch PR	Later fix	Judgment
pingdotgg#22052	pingdotgg#23886 Stop stops a thread's subagents again	The regression fix this PR is for ("Since pingdotgg#22052, the Stop button sends thread.stop…")
pingdotgg#21665	pingdotgg#22203 settings save again when the writing style has default fields	A real regression fix that batch 30 also shipped without ("pingdotgg#21665 made the writing-style setting decode into…")
pingdotgg#22052	pingdotgg#22581 idle orphaned follow-ups can be cancelled without a running parent	Related follow-up
pingdotgg#21798	pingdotgg#21799 checkpoint WAL without blocking the event loop	Upper layer of the same stack
pingdotgg#21463	pingdotgg#21464 detect replaced transcripts with large file identities	Upper layer of the same stack
pingdotgg#12567	pingdotgg#22520 keep Windows thread notifications clickable until handled	Finishes pingdotgg#12567's behavior on Windows
pingdotgg#21762	pingdotgg#23235 stop blank input placeholders from failing the run	Same fix applied to Pi
pingdotgg#14884	pingdotgg#19229 agent questions stay answerable after a server restart	Noise: mentions pingdotgg#14884 only to explain dropped tests
pingdotgg#15833	pingdotgg#21099 removing a worktree on Windows no longer deletes what its junctions point to	Noise: background citation

That is two regression fixes, five useful follow-ups and two harmless mentions, a short list to read for a 55-commit batch. Unit tests cover PR references, reverts by title and SHA, merge commits classified by PR title, references to other repositories, code fences, mirror links, and choosing the quoted sentence.

---

Written by an agent (Claude Code, claude-opus-5-5).

[github-actions]

Upstream divergence

Both sides are compared with upstream at the in-order tip fb990765a4. Excludes .repos/ and pnpm-lock.yaml.

Measure	Base	This PR	Change
Upstream files the fork changes	1,727	1,727	0
Lines changed in them	95,018	95,018	0
Upstream files the fork removes	49	49	0
Fork-only files	683	685	+2
Fork-only lines	122,674	123,067	+393

No upstream file changes how much it differs from upstream.

[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	4.9 KiB	−23 B (−0.5%)	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	−23 B (−1.9%)	2.0 KiB	✅
Codex	Live turn WebSocket decoded	20.9 KiB	20.8 KiB	−41 B (−0.2%)	29.3 KiB	✅
Codex	Live turn messages	2	1	−1 (−50.0%)	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	21.2 KiB	21.2 KiB	0 B (0.0%)	29.3 KiB	✅
Claude	Live turn messages	2	2	0 (0.0%)	8	✅

Baseline: 67439ca · PR result: 20f5d5d · 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.