The daemon's steady-state transcode ticker and a manually-launched
transcode-library backfill process were both independently claiming
pending jobs from the same queue with no shared concurrency
awareness, each capping only at its own parallelism_max. Combined
with large files easily outlasting the 30s poll interval, this
stacked to 20+ simultaneous GPU encode/decode sessions and triggered
the kernel OOM killer on a shared 15GB host running a dozen+ other
containers (confirmed via dmesg; no services were lost, no library
files were touched — the original-file-safety design held under the
crash).
Fixes: claim_pending_jobs now treats its limit as a total concurrency
cap (subtracting already-running jobs, wrapped in a BEGIN IMMEDIATE
transaction so this is correct across concurrent processes touching
the same database, not just within one). Added reset_orphaned_running_jobs,
called on real daemon startup, so a crash never permanently strands
job slots in 'running'. Lowered the default parallelism_max from 4
to 2 given the observed real-world memory pressure.
Post-grab async background swap, upgrade_locked flag to prevent the
upgrade loop from re-inflating a locally-transcoded file, and a
transcode-library CLI backfill for the existing catalog. Bitrate model
calibrated against Silicon Valley's real HEVC bitrate, scaled by
resolution and AV1's encoding efficiency. HDR/2160p and anime excluded
from this first pass. Jellyfin session polling throttles batch
encoding back to 1 stream during active playback.
qBittorrent's newer WebUI API returns 204 (not 200 "Ok.") on login success,
and a JSON success/failure summary (not plain "Ok."/"Fails." text) from
torrents/add — both broke against the currently deployed version. Also had
scan_tv_root move already-tracked episode files into their Season NN
subfolder instead of just recording wherever they already sat on disk.
