Hub uptime in /fleet freezes at the first post-boot sample (sensor rarely publishes inside the 5 s subscribe window)

SEEN 2026-09-07: after the hub OTA reboot /fleet showed uptime 3.27 s for 10+ minutes while temperatures kept updating and last_seen advanced — the uptime sensor’s state only arrives if it publishes during the short per-poll subscribe window, so the cached first value sticks. Cosmetic, but it defeats ‘did it reboot?’ checks (I initially read it as a stuck ingest). Option: take uptime from device_info / a dedicated longer subscribe, or drop the field from the fleet payload.