v0.8.3

###v0.8.3 (2026-06-01)

• Fix (docker): keep the /Docker "Check for updates" button honest. The status poll cleared the "Checking…" state on its very first 2s tick — a return inside the try fell through to a finally that reset the flag — so on a 40+ container host the button flipped back to idle almost immediately while the real per-image registry walk was still running, making the check look instant and out of sync with the container count. The poll now only concludes once the worker has genuinely completed (it has observed running:true, or a finishedAt newer than a pre-launch baseline has landed), so the button stays active for the full walk and the snapshot is re-fetched only after the worker rewrites the webui-info cache — every container's update badge reflects the fresh check.
• Fix (docker): the 60s poll limit was a hard total-runtime cap, shorter than a real 40+ container walk, so it could cut the post-check snapshot refresh off mid-walk and read stale updated flags. It is now an inactivity watchdog, reset on every observed running:true, so a provably-alive worker runs as long as it needs while a worker that never establishes (or wedges) its lock still bails. New checkUpdatesCompleted() pure helper covers the decision; +5 unit cases. 456 TS suites green; backend untouched (verified correct against the stock getAllInfo/reloadUpdateStatus path on a live host).