Since this issue is known, I leave my findings here. Seems nchan can be a bit diffcult sometimes. I killed (kill -9) the main process and restarted it, using rc.nginx which worked.
Unraid 6.12.3.
Observed a weird behavior of docker...
I use the DockerClient class for starting and stopping containers for the appdata.backup. Now, after a few days with my new setup server, I noticed, that my guacd container stays off after backup. The log says, that it could not be stopped, because "Code: " (no further code). This is the result straight out of stopContainer function.
So I made a test:
DockerClient->stopContainer always fails to stop guacd and hits the timeout. If I use "docker stop guacd" in the console, its stopping as wanted.
Did anyone noticed this as well? I dont know whats the difference, both commands should talk to docker via the socket, right?
Wait.. Just while writing this... I forgot to specify the -t parameter for "docker stop guacd". As soon as I specify -t, it does the same.
With this knowledge, I solved my own problem during writing this thread (I still leave it here). docker stop hits the default 10 second timeout and KILLS the container without changing the output (just displays the container name, as usual). Obviously its something within guacd which ignores the SIGTERM.
But the DockerClient->stopContainer returns an error if the KILL happened. I know that this is coming straight out the docker API. So I have to catch that case with a test.
Rebooting again after the cache drive removal fixed the issue, maybe worth noting.
So something between "Yes, I want to do this" (Remove cache and start Array) and the next reboot, seems to prevent creating a new system share.
KluthR's post in nginx: nchan_store_init_worker: Assertion `procslot_found == 1' failed was marked as the answer