Rackmount myServer - whole server unresponsive

Well it's been up now since 21st December which is longer than it was managing before. So I assume something I pulled off the server was the culprit - zigbee2mqtt or nodered presumably. Shame as you make provision for zigbee2mqtt in the /home/applications directory and nodered isn't too heavyweight. Anyway now it seems to be stable I'll leave it as it is and just live with having stuff scattered around a bit. Stability is more important than a bit of complexity.

And the server died again last night at 00:18, exactly the same symptoms. Just when I was thinking it was fixed. Bah.

I've now started the myServer Pi sending syslog messages to my Synology NAS. Maybe something will show up in there when it spontaneously reboots.

It reboots on its own? Or it becomes unresponsive? Or it shuts down?
What does HTOP say?
What is in the log (we don't need everything, just things of interest posted)?
I gotta assume it's some custom script that isn't cleaning up after itself consuming memory.
Are you keeping chrome running on the desktop? Typically that is the largest memory hog. myServer should run "headless" and no chrome instance running on the hardware.
We have systems running years without a reboot. And no hardware failures that weren't due to power cycling. So, with that background, we believe it's something unique to your system. So, look at those uniqueness's for hints.

I've posted details and logs before. It looks like either the whole server spontaneously reboots or myServer spontaneously restarts. It's been impossible to tell because whatever happens the Pi ends up in a completely unresponsive state - video out still on showing the desktop, but the clock in the bottom right corner showing the time the problem occurred and physical keyboard and mouse don't work. Pinging the server gets no response as well - it's fully dead apart from the frozen video output. myServer log shows it initializing but then the log goes silent until I yank the power and everything restarts. At which point the Debian logs e.g. dmesg get overwritten with the new boot hence why I can't tell if the whole Pi restarted or just myServer.

And that's why I'm now sending syslog to an external server - that'll be preserved through a restart so I'm hoping I might get more info on what's going on.

And yes it's probably something on my Pi but as I've removed extraneous software it should only be myServer and its drivers running. I'll probably end up re-imaging and reconfiguring but I've been hoping to avoid that.

PS as a headless server I don't understand why you run a whole desktop installation. Seems unnecessary.