This is a clone of issue OCPBUGS-19430. The following is the description of the original issue:
—
This is a clone of issue OCPBUGS-18772. The following is the description of the original issue:
—
MCO installs resolve-prepender NetworkManager script on the nodes. In order to find out node details it needs to pull baremetalRuntimeCfgImage. However, this image needs to be pulled just the first time, in the followup attempts this script just verifies that this image is available.

This is not desirable in situations where mirror / quay are not available or having a temporary problem - these kind of issues should not prevent the node from starting kubelet. During certificate rotation testing I noticed that the node with a significant time skew won't start kubelet, as it tries to pull baremetalRuntimeCfgImage for kubelet to start - but the image is already on the nodes and it doesn't need refreshing.