The pinned images functionality is not working
    

IPI on AWS version:
$ oc get clusterversion
NAME      VERSION                              AVAILABLE   PROGRESSING   SINCE   STATUS
version   4.18.0-0.nightly-2024-10-28-052434   True        False         6h46m   Cluster version is 4.18.0-0.nightly-2024-10-28-052434

    

Always
    

    1. Enable techpreview
    2.  Create a pinnedimagesets resource

$ oc create -f - << EOF
apiVersion: machineconfiguration.openshift.io/v1alpha1
kind: PinnedImageSet
metadata:
  labels:
    machineconfiguration.openshift.io/role: worker
  name: tc-73623-worker-pinned-images
spec:
  pinnedImages:
  - name: "quay.io/openshifttest/busybox@sha256:0415f56ccc05526f2af5a7ae8654baec97d4a614f24736e8eef41a4591f08019"
  - name: quay.io/openshifttest/alpine@sha256:be92b18a369e989a6e86ac840b7f23ce0052467de551b064796d67280dfa06d5
EOF

    

The images are not pinned and the pool is degraded

We can see these logs in the MCDs

 I1028 14:26:32.514096    2341 pinned_image_set.go:304] Reconciling pinned image set: tc-73623-worker-pinned-images: generation: 1
E1028 14:26:32.514183    2341 pinned_image_set.go:240] failed to get image status for "quay.io/openshifttest/busybox@sha256:0415f56ccc05526f2af5a7ae8654baec97d4a614f24736e8eef41a4591f08019": rpc error: code = Unavailable desc = name resolver error: produced zero addresses

And we can see the machineconfignodes resources reporting pinnedimagesets degradation:

  - lastTransitionTime: "2024-10-28T14:27:58Z"
    message: 'failed to get image status for "quay.io/openshifttest/busybox@sha256:0415f56ccc05526f2af5a7ae8654baec97d4a614f24736e8eef41a4591f08019":
      rpc error: code = Unavailable desc = name resolver error: produced zero addresses'
    reason: PrefetchFailed
    status: "True"
    type: PinnedImageSetsDegraded

    

The images should be pinned without errors.

    

Slack conversation: https://redhat-internal.slack.com/archives/C02CZNQHGN8/p1730125766377509

This is Sam's guess (thank you [~sbatschelet] for your quick help, I really appreciate it):
My guess is that it is related to https://github.com/openshift/machine-config-operator/pull/4629
Specifically the changes to pkg/daemon/cri/cri.go where we swapped out DialContext for NewClient. Per docs.
One subtle difference between NewClient and Dial and DialContext is that the former uses "dns" as the default name resolver, while the latter use "passthrough" for backward compatibility. This distinction should not matter to most users, but could matter to legacy users that specify a custom dialer and expect it to receive the target string directly.