As @truecharts says, this is bad advice, but I recently switched to cobia and even after stopping the k3s service my docker containers were terminated, which is probably the desired behavior. I could work around that by editing /usr/local/bin/k3s-kill.sh and commenting out the line that says "killtree $({ set +x; } 2>/dev/null; getshims; set -x)". As for nextcloud, I just added my own cronjob in the truenas gui. I get 1-5 % cpu usage with about 17 running applications (idle, of course). But again this is probably bad advice!!
Edit: the changes to /usr/local/bin/k3s-kill.sh probably get reset after updating your truenas install, so keep that in mind
Edit: I wrote a little shell script that should fix this (including cron etc.), please read the whole README before doing anything:
https://github.com/sdaqo/k3s-killer
