When reduce_storage_consumption() encountered a FileNotFoundError
(file deleted outside Frigate), it silently skipped the recording
without removing it from the database. Over time this caused the DB
to accumulate stale entries, making "Frigate recordings tracked" in
/system#storage dramatically overstate actual disk usage.
The bug also affected cleanup behaviour: stale entries don't count
toward freed-space accounting, so Phase 2 (force-delete retained
recordings) could trigger prematurely when most old entries were stale.
Fix: always append the recording to deleted_recordings regardless of
whether the file existed, so the DB entry is removed. freed-space
accounting is unchanged — FileNotFoundError still does not increment
deleted_segments_size since no actual disk space was recovered.
Applied to both Phase 1 (non-retained) and Phase 2 (retained) loops
inside reduce_storage_consumption().
https://claude.ai/code/session_01DMdSSQhQfTuXmzPtRvJmLB
f528a16065