There's a possible flake in openshift-tests because the upstream test "Should recreate evicted statefulset" occasionally causes kubelet to emit a "failed to bind hostport" event because it tries to recreate a deleted pod too quickly, and this gets flagged by openshift-tests as a bad thing, even though it's not (because it retries and succeeds).

I filed a PR to fix this a long time ago, it just needs review.