https://bugzilla.redhat.com/show_bug.cgi?id=1943334#c4

During upgrades and on startup, we need a way to see when ovnkube or sdn pods go into ready state. Until then we shouldn't allow scheduling on the nodes since they wouldn't work. A solution of using taints on nodes was explored but that did not work well since components depending on host-networked pods would have to tolerate that taint.

A better approach is to implement this in CRI-O side. This card will track this RFE since the implementation needs some design discussions.