[andrewcai8]

Merges 134 upstream commits, through upstream's newest commit with green CI. Notable ones: Claude background agents settle when their process exits (pingdotgg#19464), Stop no longer leaves a killed background command showing as running (pingdotgg#18163), forking a Claude thread works on instances with their own home (pingdotgg#18662), and native subscription accounts are deduplicated in usage (pingdotgg#10700).

Conflicts:

• Orchestrator.ts: took upstream's Stop reconciliation. The stopped turn's own provider thread goes through stillReportsBackgroundWork, which covers the failed-turn case the fork's fix(server): background jobs stay stoppable after a turn fails on a usage limit #193 fixed. Any other provider thread with a live session keeps its roster, as upstream's BackgroundWorkStop tests expect, so the fork's liveSessionRunsWork is removed. fix(server): background jobs stay stoppable after a turn fails on a usage limit #193's tests still pass.
• ThreadLaunchService.ts: took upstream's Scratch normalization of the launch input. The fork's cloud-only refusal runs on the normalized input, before any thread is created.
• ScheduledTaskService.ts: kept the fork's target beside upstream's normalized workspaceStrategy.
• Web:
  • BranchToolbar and ChatView keep the fork's cloud Run-on picker with upstream's optional onEnvironmentChange.
  • A held cloud send keeps its draft and isn't marked promoted until it is sent.
  • Sidebar takes upstream's routeThreadRef beside the fork's homeThreadRef.
  • The usage page says "to now" with upstream's 12/24-hour setting.
• deploy-relay.yml keeps the fork's opt-in deploy.

Follow-ups the merge needed:

• DesktopWindow.ts had the same DesktopState import and lookup added on both sides, so the duplicates are removed.
• ScheduledTaskService.cloud.test.ts provides ManagedProjectFolders, which the service now needs.

Verification:

• Typecheck is clean for contracts, shared, client-runtime, provider-core, server, web, mobile and desktop.
• Repo lint has no errors.
• Server tests for the files both sides changed, plus the scheduled task, cloud fleet, account switch and fork migration tests, pass (18 files, 848 tests).
• Web unit tests pass (7226).
• Mobile unit tests pass apart from the two macOS-only native race tests in scripts/, which are unchanged from upstream.

🤖 Generated with Claude 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	4.9 KiB	−30 B (−0.6%)	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	−30 B (−2.5%)	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	−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: 8464420 · PR result: 4194cb5 · 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.