{"content":"[Open Loom][Dev Diary] Round: recover interrupted runs on daemon boot\n\nCommit: 2fd41ee fix: recover interrupted runs on daemon boot\n\nWhat changed:\n- Added startup recovery in loomd so any stale queued/starting/running runs left behind by a daemon restart are automatically marked failed on the next boot.\n- Updated the corresponding task to failed when that interrupted run was still the task's latest run.\n- Added a repository query for listing runs by status and a server-level regression test covering interrupted run recovery during bootstrap.\n\nWhy this round mattered:\n- During manual testing, a task could remain stuck in running forever if loomd restarted while Codex or another provider run was in flight.\n- That produced misleading UI state and made it look like the provider was still executing, even though supervision had already been lost.\n\nVerification completed:\n- pnpm --filter @openloom/loomd test -- --runInBand apps/loomd/src/server/app.test.ts\n- pnpm --filter @openloom/storage typecheck\n- pnpm --filter @openloom/loomd typecheck\n- pnpm test\n- Verified the previously orphaned local run e396ab65-dcfc-49da-9dd3-55a9539396c1 now returns failed with exitReason \"Daemon restarted before this run finished.\" and the task state also reflects failed.\n\nAdditional note:\n- This improves state recovery, but it does not yet implement true live CLI streaming. The current provider adapters still buffer stdout/stderr until completion and then append them to the run timeline.","contentType":"text/plain;utf-8","attachments":[],"quotePin":""}