fix: apply persisted active db before startup health

fix: restore active db from persisted state on startup

fix(infra): paginate BM25 fetch + volume-preserve guard against clobber

fix: remaining infra + extraction issues from session audit

fix: never publish empty DB dirs over release backups (husk guard)

fix: republish missing allowed dbs on startup

fix: honor deleted db tombstones on startup restore

fix: hard-lock db deletion to explicit admin action

refactor: harden crawl chunking and restore lock

fix(multitenant): restore missing tenant db from release on widget request

fix: remove stale /dev/shm sync code from upload path

fix: sync /dev/shm back to disk before zip upload

fix: stream zip extraction to avoid OOM on large files + fix zf->z bug

fix: stale .tmp_sync files causing permanent 'being restored' state

fix: merge operational config fields on HF restart + agentfactory scope fix

fix(sync): persist cleared DBs as empty on GitHub

fix(sync): load local DB when GitHub sync fails (401/expired PAT)

fix(github_sync): always keep repo config.json over zip version

fix: preserve rich config.json during GitHub sync; fix eval OOS filter for retail DBs

evals: avoid HF sync deadlocks during owner generation

refactor: extract services/github_sync.py (session 3)