MicroShift fails to start because kubelet cannot talk to crio over a socket because crio is SIGABRT'ed by systemd.

CRIO complains for a minute that it "Failed to get network for name: ovn-kubernetes". Then following is logged:

msg="Will not notify watchdog because CRI-O is unhealthy: health checker failed: runtime status \"NetworkReady\" is invalid: Network plugin returns error: no CNI configu
ration file in /etc/cni/net.d/. Has your network provider started? (reason: NetworkPluginNotReady)" file="watchdog/watchdog.go:64"

systemd[1]: crio.service: Watchdog timeout (limit 1min)!
systemd[1]: crio.service: Killing process 22355 (crio) with signal SIGABRT.

crio 1.32

Everytime with 1.32 in our CI

    1. Install MicroShift, install crio 1.32 (from the ocp mirror).
    2. `systemctl start microshift`
    3.
    

    Starting microshift fails (it eventually starts when everything catches up to, but first start is important in CI).

    crio isn't killed because the CNI is not ready yet. MicroShift starts because kubelet can contact crio.

    Example journalctl (search for SIGABRT): https://gcsweb-ci.apps.ci.l2s4.p1.openshiftapps.com/gcs/test-platform-results/logs/periodic-ci-openshift-microshift-release-4.19-periodics-e2e-aws-tests-bootc-arm-nightly/1878623192738697216/artifacts/e2e-aws-tests-bootc-arm-nightly/openshift-microshift-e2e-metal-tests/artifacts/scenario-info/el95-src@isolated-net/vms/host1/sos/journal_2025-01-13_02:58:46.log

Here's crio trace log from another machine:
https://drive.google.com/file/d/1sQHFzV_cJOPMBkbItBIocusPtWKSyjp1/view?usp=sharing