This is a clone of issue OCPBUGS-56445. The following is the description of the original issue:
—
This is a clone of issue OCPBUGS-56211. The following is the description of the original issue:
—
This was noticed during the upgrade(4.19.ec-4 to 4.19.ec-5) of a build cluster in tech preview mode(slack thread). The boot image controller(MSBIC) currently waits for at least one master node to be updated with a timeout of 15 minutes, which is currently signaled by the golden configmap recording the current MCO hash and release version. This was originally put in place so that the MCO does not update the boot images before the new MCO images have been rolled out, which may result in "backward" pivots, if a node is scaled during an upgrade.

During the upgrade of the build cluster, a CRD apply failed(unrelated to the boot image updates), causing the MSBIC to timeout before even the new MCO controllers were rolled out. The MCO's operator pod is responsible for rolling out the new controller pods, however due to the degrade set by the MSBIC, this did not take place, and the operator was stuck in a never ending loop. 

To rectify this, instead of waiting up to 15 minutes, the MSBIC will non fatally exit the sync and attempt a follow-up sync when the golden configmap is updated. This will still result in the boot image being updated, but will not cause issues when the initial parts of the cluster upgrade was slow.