We currently have those set of steps in our recovery procedure:

• Correct cluster-split-avoiding shutdown procedure (OCPBUGS-36218)
  • Which also needs to work in network split scenarios
• Avoiding interferences with keepalived
• Removal of any VIPs (related to keepalived)
• PROXY handling
• kubelet PKI removal
• csr approval
• disable CEO quorum guard
• (OVN deletion + machine removal, we know we can skip with rev bumping)

With rev bumping enabled, which ones can we skip?

Note that most of those steps apply to bare metal deployments, so we might also try to run the disaster recovery tests on a baremetal workflow.