v0.8.5

###v0.8.5 (2026-06-02)

• Fix (docker): stop /Docker from blanking — and stop a stale "update available" badge — while a container update is in flight. During the stock update_container flow's rewrite of the webui-info docker.json, getAllInfo() momentarily returns an empty list (still HTTP 200), which overwrote the populated store with nothing and poisoned the per-tab snapshot cache, leaving the page blank across refreshes until the update finished. docker-state.php now cross-checks getAllInfo() against getDockerContainers() (the docker socket, unaffected by the rewrite) and responds 503 when the daemon still lists containers but the cache reads empty; the client keeps its current rows and retries on the next resync instead of clobbering them. A genuinely empty server (every container deleted) still returns [] and renders the empty state — fixing the inverse "ghost rows after delete-all" bug the first client-side guard introduced. No extra docker calls: both sources were already fetched on every load.
• Fix (docker): clear the freshly-updated "update available" badge without a manual reload. The _DONE_ handler now resyncs before clearing the update probes, so the id-rotation reconcile (the definitive "pull succeeded, container recreated" signal) can force-clear the badge over Unraid's lagging digest cache while the probe still exists; clearing first left reconcile nothing to match and the stale badge persisted. New docker-state.test.php for the transient-empty disambiguation plus boot.ts/store.ts ordering coverage; 116 TS + PHP suites green.