We are pushing to find a resolution for OCPBUGS-11591 and the SDN team has identified a key message that appears related in the system journald logs:

Apr 12 11:53:51.395838 ci-op-xs3rnrtc-2d4c7-4mhm7-worker-b-dwc7w ovs-vswitchd[1124]: ovs|00002|timeval(urcu4)|WARN|Unreasonably long 109127ms poll interval (0ms user, 0ms system)

We should detect this in origin and create an interval so it can be charted in the timelines, as well as a unit test that fails if detected so we can see where it's happening.