I've been chasing an intermittent issue where Home Assistant stops receiving Maker API events from my Hubitat hub, and after extensive testing, I've isolated it to the Home Assistant / integration side rather than Hubitat or networking. Posting here in case others have hit this and found a permanent fix.
Environment:
Home Assistant is running in Docker Desktop on Windows (not native Linux/docker-ce). Docker Desktop on Windows runs containers inside a hidden Hyper-V/WSL2 VM with a network proxy layer (vpnkit/com.docker.backend.exe), rather than true host networking like Linux Docker has. This is relevant because it means an external "is this port open" check only confirms Docker Desktop's proxy is forwarding traffic into the container — it does not confirm that the Home Assistant process itself is still actually processing that traffic.
Symptom: Periodically (every few days), device state updates from Hubitat stop arriving in Home Assistant. Hubitat's local control and UI continue working normally. Commands sent from HA to Hubitat still work. Only inbound event updates (Hubitat → HA) stop.
What did NOT fix it, despite repeated testing:
Restarting the Home Assistant Docker container
Reloading the Hubitat integration from HA's integration page
Rebooting the Hubitat hub (soft reboot via Settings → Reboot)
Fully power-cycling Hubitat for 30+ seconds
Switching Hubitat between static IP and DHCP
Router/ARP-layer fixes (static ARP entries, ARP cache tunables, manual ARP flush, gratuitous ARP broadcasts) — confirmed via ping that basic IP reachability between Hubitat and HA was never actually broken during the outage
What DID fix it:
A full Home Assistant application restart (Settings → System → Restart, not just restarting the Docker container) reliably restored event flow immediately.
My working theory: My autoheal/healthcheck setup was monitoring the listening port for Hubitat events and reported it healthy throughout the outage, yet events still weren't being processed. Given the Windows Docker Desktop networking layer described above, I suspect this healthcheck was only confirming Docker Desktop's proxy was forwarding connections — not that HA's own event loop or aiohttp handler behind it was still functioning. Something inside the integration or HA core's event handling appears to get stuck in a state where the listener stays technically reachable but stops actually parsing/dispatching incoming Hubitat POSTs. An integration reload doesn't clear it, but a full app restart does — suggesting the stuck state lives above the integration layer, possibly in HA core's event bus or aiohttp handling, and is masked by the extra Windows networking abstraction layer.
Questions for the community:
Has anyone else seen the Hubitat integration (or Maker API webhook listener generally) go into a "port open but not processing" state like this?
Has anyone experienced this specifically on Docker Desktop for Windows vs. native Linux Docker — could the Hyper-V/WSL2 proxy layer be masking or contributing to this?
Are there specific HA core or integration debug logs worth enabling to catch this in the act next time (rather than just restarting immediately)?
Is there a known HA core issue related to webhook/event listeners silently stalling that this might match?
Happy to share HA logs from the next occurrence if it'll help debug this further.