Based on the information provided:
I hypothesize that certain combinations of workloads trigger the (configurable?) conditions
[quote="wtempel, post:23, topic:12832"]
`> 50.00% for > 20s with reclaim activity`
[/quote]
for systemd-oomd to `kill` processes, including some CryoSPARC processes in a way that prevents those processes from "cleaning up after themselves". This _could_ explain the presence of an orphaned
/tmp/cryosparc-supervisor-206773da3c7c06e952eddaffaea9188d.sock
file. Before removing the file, please confirm that the corresponding supervisord process is in fact no longer running (related discussion).
A google.com search indicates that some consider systemd-oomd as aggressive and suggest a variety of interventions. May I suggest a consultation with your system administrator about a potential reconfiguration, and the avoidance of workload patterns that exhaust available system RAM.