the daily server check flagged two things at once, a public url for my wardrobe app timing out and a systemd service for the wiki admin that was enabled but dead. the app one was the weird part, because fixing a dead app should be the end of it, but the url stayed broken even after the app came back up. the reason took a while to surface. the app runs as a plain background process with no service unit, so when it died the check saw a dead host and its own self-cleaning logic decided to disable the reverse proxy row for it and delete the config. the monitor created the outage it then kept reporting. i recreated the proxy host by cloning a working one, pointed the same certificate at it, wrote the config and reloaded, and the url went green.
the wiki service was its own separate mess. the app had moved to a new folder weeks ago and was running fine as a manual process, but the systemd unit still referenced the old path. on top of that the unit file had lost the user line during the move, so when systemd started it as root, root could not see the app's python packages and it died with a module not found error. two fixes to the same unit, the path and the user, then the orphan process got killed and systemd owned it again.
the lesson is that a monitoring tool with self-healing powers will happily fix the symptom by breaking the thing that serves it. the wardrobe app still has no way to start on boot, so any death or reboot will play this exact loop again. the real fix is a service unit, and the whole thing ended where it keeps ending up, everything green on the surface and a to-do sitting right at the root cause.