A flaw was found in cri-o, as a result of all pod-related processes being
placed in the same memory cgroup. This can result in container management
(conmon) processes being killed if a workload process triggers an
out-of-memory (OOM) condition for the cgroup. An attacker could abuse this
flaw to get host network access on an cri-o host.
It's not clear to me what version this was fixed in, if at all.