[yordis]

• Upstream now meters checkpoint git attempts and finds the resource monitor from the built server, so their versions replace the fork's equivalent fixes rather than carrying two implementations.
• Telemetry resource definitions keep the fork's shape because upstream took the t3code. keys but still emits service.runtime and lacks instance ids and browser attributes; ledger 0026 is narrowed to what upstream has not taken.
• The observability guide keeps the fork's fuller wording and picks up upstream's note on the t3code. prefix so neither side's guidance is lost.
• The lockfile takes upstream and is regenerated so the fork's own dependency bumps and list patch stay locked.

[cursor]

PR Summary

Medium Risk
Desktop quit and CDP screenshot lifecycle changes affect process exit and preview stability; mobile approval-send and scroll/header inset changes touch core thread and navigation UX.

Overview
Adds a scheduled GitHub Action that closes issues labeled needs more info after seven days without a reply.

Desktop quit handling is reworked around a QuitState phase machine: the first quit runs shutdown and waits, and only after a 10s stall does a repeated Cmd-Q or SIGTERM call exit(0) so a double-quit cannot orphan the backend mid-stop. Preview CDP screenshots are serialized per guest tab (with queue drain on navigation, detach, or relay release) to avoid Chromium renderer crashes from overlapping captures. PDF guest pages get a transparent input layer so annotation tools work over the built-in viewer.

Mobile gains host-backed PR label/author filter search, colored label dots in filter menus, a Load more control when pagination is short, and scroll/header behavior that relies on UIKit automatic content insets plus optional header minimization on scroll-up on Duo devices (replacing manual useRestingTopOffset). Threads can send approval feedback instead of only decline-and-redirect when the pending approval supports it. Composer and feed context/thread chips get contrast-safe accents, project favicons/monograms in the command palette, rasterized project glyph images, and richer thread link chips. Android adds a swipe-dismissible thread undo toast; several unused thread-activity inspector/navigation helpers are removed. Desktop observability tests assert unpackaged runs report the app package version; react-native-screens is bumped to a new fork tarball.

^{Reviewed by Cursor Bugbot for commit 82c149b. Bugbot is set up for automated code reviews on this repo. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes using default effort and found 2 potential issues.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, have a team admin enable autofix in the Cursor dashboard.}

^{Reviewed by Cursor Bugbot for commit 82c149b. Configure here.}

[cursor]

Workspace chip can hide branch picker

Medium Severity

The workspace ComposerInlineControl now uses maxWidth of 100% and is not wrapped in min-w-0 shrink, unlike the branch control. A long workspace label can take the full row, so the branch picker shrinks away and cannot be opened.

 

^{Reviewed by Cursor Bugbot for commit 82c149b. Configure here.}

[cursor]

Approval feedback failures fail silently

Medium Severity

When the first pending approval has supportsFeedback, send always takes the feedback path and returns false if sendApprovalFeedback fails. That happens when the request is not live or the respond call does not succeed. The composer then aborts without declining other approvals, sending a redirect, or showing an error.

 

^{Reviewed by Cursor Bugbot for commit 82c149b. Configure here.}

[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	20.9 KiB	20.9 KiB	0 B (0.0%)	29.3 KiB	✅
Codex	Live turn messages	2	2	0 (0.0%)	8	✅
Claude	Total thread wire	5.0 KiB	5.0 KiB	+7 B (+0.1%)	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	+7 B (+0.6%)	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: 17f7c43 · PR result: 82c149b · 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.