Pre-flight checks should check if a major OpenShift upgrade results in API removal. The upgrade should not be allowed until all resources and assets have been removed from the cluster. This will prevent:

• unexpected user disruption (e.g. DeploymentConfig workloads)
• orphaned data being left in etcd

We should also explore adding logic similar to RemoveUnavailableKinds in kube-apiserver to the extension apiservers (openshift-apiserver and oauth-apiserver). Alternatively, we could simply remove the type registration in the extension apiservers. Or both.

Resource removal will usually be done alongside removing or rewriting all the existing tests that reference these types.

https://redhat-internal.slack.com/archives/C0A475V81LN/p1769091023061799 for context