Fix ruststreamer recordings across service restarts
Urgently diagnose and fix the regression where stopping a ruststreamer meeting processes only the recording session created by the latest service process instead of all UUID recording directories accumulated across restarts. Evidence: each recording directory has a durable `.recording-session.json` manifest. Manual processing discovers these manifests, but automatic stop-job creation only captures `PassThrough.recording_sessions`, which is in-memory and resets when the service restarts. Acceptance criteria: - Automatic stop processing includes all durable recording sessions for the exact stream, including sessions created by earlier service processes. - Current in-memory sessions are retained and UUIDs are deduplicated. - The complete cross-restart session set is grouped and ordered for concatenation/upload. - A regression test covers durable earlier sessions plus the current session. - Focused ruststreamer tests and workspace formatting/checks pass.
Checklist
5 of 5 complete- Automatic stop jobs merge durable pre-restart sessions with current sessions and deduplicate UUIDs
- Regression test verifies all cross-restart sessions are ordered into the processing job
- Ruststreamer tests, workspace formatting, and workspace check pass
- Historical shutdown-finalization failures do not permanently poison valid recording sessions
- Processing-job creation errors cannot prevent the live stream from shutting down
- Area
- Streaming
- Project
- streaming-engines
- Kind
- Bug
- Priority
- High
- Task ID
tsk_f8jgbsm99g- Updated