BREAKING CHANGE
- the container now runs as uid 1000 instead of root, and
the /app/.oauth mount point is gone. Volumes created by an earlier
release contain root-owned files; auth/storage.py chmods tokens.db on
startup, which raises PermissionError for a non-owner, so a stale volume
fails to start rather than degrading. Recreate dev volumes with
docker compose down -v, orchown -R 1000:0a persisted host
directory. Any /app/.oauth mount should simply be dropped.
Feat
- drop the dead /app/.oauth EFS mount from the ECS task
- run the container as uid 1000 with the venv outside /app
Fix
- set HOME=/tmp so the parsing path works as uid 1000
Refactor
- run as a real appuser account instead of a bare numeric uid