Fixed
- A re-scan whose TMDB lookup fails no longer re-drifts an API-matched title's tracks (or wipes manual edits). When the fresh lookup times out, the scan classifies against the heuristic native (the first audio track) while the upsert preserves the stored authoritative native — so a Chinese dub got re-kept on an English show and manual track edits were lost. The persist path now keeps the stored tracks when the stream layout is unchanged, and the one-time authoritative-native heal re-runs once to fix already-drifted rows.
Docker images
Pinned to this release:
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.112 # portable (amd64 + arm64, CPU)
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.112-nvenc # x86_64 + NVIDIA GPU (NVENC)
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.112-edge # as above but ffmpeg master
docker pull ghcr.io/i-ial9000/shrinkerr:v0.9.112-edge-nvencFloating tags (:latest, :edge, :nvenc, :edge-nvenc) track the most recent release of each lineage.