PR 24362 runtime verification, 2026-11-01 (UTC)

Commands: git worktree add ../main github/main; timeout 600 vp i
Main: T3CODE_DEV_SEED=0 setsid nohup vp run dev > /tmp/main-dev.log 2>&1 &
PR: git switch --detach 47c1c0efeacd42653cb8a2476b24eb3d42444236; timeout 600 vp i
PR server: T3CODE_DEV_SEED=0 setsid nohup vp run dev --home-dir /home/user/t3code/.t3 > /tmp/dev.log 2>&1 &
Main was stopped before starting PR. Ports read from dev-runner: main 8693/16733; PR 5733/13773.
Browser: repo Playwright 1.60.0, /usr/bin/google-chrome, real web client.
UI-created project: /home/user/t3code, a real git repository. UI-created thread on each server.
Real provider: Codex gpt-6.1-sol through local CLIProxyAPI. No Replay provider, mock server, or seeded data.
Network fault: Playwright routeWebSocket forwards actual client/server frames unchanged, except it withholds one actual successful server.reportClientActivity reply. The socket stays open; no fake reply is generated.

github/main 5732bbee397f0a2b0e61d0c2e432f01dca6df3dc
fault request 2026-11-01T03:05:40.300Z; withheld reply 2026-11-01T03:05:41.496Z
observed 83.244 seconds after fault; 17 real Pong frames
2026-11-01T03:05:40.300Z -> activity request id=25, ttl=45000, scopes=provider-status,vcs-status
UI showed provider responses MAIN_ACTIVITY_OK and NEARBY_ACTIVITY_OK.

PR #24362 47c1c0efeacd42653cb8a2476b24eb3d42444236
fault request 2026-11-01T03:09:20.513Z; withheld reply 2026-11-01T03:09:20.519Z
observed 75.222 seconds after fault; 20 real Pong frames
2026-11-01T03:08:57.010Z -> activity request id=20, ttl=45000, scopes=provider-status,vcs-status
2026-11-01T03:09:20.513Z -> activity request id=22, ttl=45000, scopes=provider-status,vcs-status
2026-11-01T03:09:41.685Z -> activity request id=24, ttl=45000, scopes=provider-status,vcs-status
2026-11-01T03:09:45.512Z -> activity request id=28, ttl=45000, scopes=provider-status,vcs-status
2026-11-01T03:10:10.514Z -> activity request id=29, ttl=45000, scopes=provider-status,vcs-status
2026-11-01T03:10:35.524Z -> activity request id=30, ttl=45000, scopes=provider-status,vcs-status
2026-11-01T03:09:35.514Z -> Interrupt requestId=22 (15.001s after unanswered request)
2026-11-01T03:08:57.148Z <- activity reply requestId=20, Success
2026-11-01T03:09:41.801Z <- activity reply requestId=24, Success
2026-11-01T03:09:45.522Z <- activity reply requestId=28, Success
2026-11-01T03:10:10.523Z <- activity reply requestId=29, Success
2026-11-01T03:10:35.549Z <- activity reply requestId=30, Success
server.getBackgroundPolicy: activeForegroundLeaseCount=1, activeScopeKeys=provider-status,vcs-status:/home/user/t3code
lease updatedAt=2026-11-01T03:10:35.534Z, expiresAt=2026-11-01T03:11:20.534Z
snapshot observedAt=2026-11-01T03:10:35.825Z
UI showed provider responses MAIN_ACTIVITY_OK and NEARBY_ACTIVITY_OK.

Neighboring behavior: real Codex follow-up completed during the fault/recovery window.
Fresh browser context reopened same thread with saved auth and both messages present.
2026-11-01T03:11:28.512Z -> activity report after reconnect id=22
Reconnect policy: activeForegroundLeaseCount=1

blacksmith-check: installed dependencies; vp test run apps/web/src/lib/backgroundActivityReporter.test.ts apps/mobile/src/connection/background-activity.test.ts: 2 files, 7 tests passed.
GitHub CI: all required checks pass on exact PR head; no source changes or pushes needed.
Not exercised: native mobile/Electron, macOS/Windows, remote/relay/tunnel, other providers, or a naturally occurring server hang. Fault was induced at the actual network boundary.