Fixed
- "Refresh metadata" (heuristic → API) no longer aborts on the first DB-lock and now completes. A batch write that hit "database is locked" under write contention (conversions + a scan) propagated out and killed the entire refresh, leaving most heuristic rows — much of the TV library — never upgraded to API-sourced language. Batch writes now retry on a transient lock and defer only the failed batch instead of aborting the whole run.
Docker images
Pinned to this release:
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.63 # portable (amd64 + arm64, CPU)
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.63-nvenc # x86_64 + NVIDIA GPU (NVENC)
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.63-edge # as above but ffmpeg master
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.63-edge-nvencFloating tags (:latest, :edge, :nvenc, :edge-nvenc) track the most recent release of each lineage.