Summary
Deleting the only active job directly in a download client could leave a stale tracking record in Listenarr. Every later grab of the same audiobook returned HTTP 409. The download monitor now clears confirmed orphans after the five-minute grace period so the audiobook can be grabbed again.
Changes
Added
- Added Playwright coverage for re-grabbing after deletion and for Activity recovery after a real qBittorrent outage.
Changed
- The download monitor checks a complete live queue and looks up tracked job IDs independently of category across qBittorrent, Transmission, SABnzbd, and NZBGet. A successful empty queue can trigger cleanup after the job lookup confirms absence.
Fixed
- Preserve tracked records when the client is unavailable or its response cannot establish whether the job is present. Guard cleanup and telemetry writes against concurrent import and tracking updates.
- Activity reports unavailable clients and hides their stale rows while continuing to show downloads from healthy clients.
Activity during a qBittorrent outage (browser fixture with a healthy Transmission download).
Testing
- The Run Tests workflow passed on
7618f74c, including Linux and Windows backend tests and all 32 Playwright E2E tests. - Live qBittorrent tests verified that deleting the sole torrent allows a re-grab, and that an outage hides the row until the original torrent returns. Active duplicate grabs returned HTTP 409.
- Earlier live SABnzbd tests verified that older tracked jobs survive exact-ID history checks and that missing or malformed history data cannot trigger cleanup without confirming absence.
Notes
Continues #755 and preserves the original commits by @t3chnaztea.
Automated Canary build
