Yeah, none of you are crazy! Sorry about this, I am looking into it now.
Update: The “ad-hoc” check uses MQTT and the client had a bug so it wouldn’t report its online status after reconnecting. The MQTT server was restarted a few days ago to move to another node in the cluster. I wlll add monitoring for “too few monitors connected via mqtt” and also improve the error handling on the website. I’ve fixed the client, but not tagged it in a release yet so in the meantime I just restarted a few of the monitoring clients.