Description of problem:

I had an installation of SNO (version 12.5) almost fail because the pods in openshift-marketplace were not deploying successfully .

The cluster went through the following events>

• image registry not available first
• it started trying to deploy the 5 pods in openshift-marketplace namespace

Unluckily, with the exception of the marketplace operator, the rest of the pods don't belong to deployment or replicas (or there isn't anything else controlling them), given a copy of each was failing, it started creating multiple copies at the same time of the pods (up to 60 pods in total). In particular of:

• certified-operators
• community-operators
• redhat-marketplace
• redhat-operators

I had to eventually create a project pod quota to limit the number of duplicated pods, only then, and with the registry available again, I was able to deploy it successfully.
The duplicated pods where consuming all possible resources of the cluster, leaving none to the registry.

Version-Release number of selected component (if applicable):
OCP12.5

How reproducible:
Unsure but killing the registry during the process of installation might do the trick

Expected results:
Having a limit on number of copies generated of a pod inside marketplace namespace (or in general at installation time for any infra component)