DoD:

Consider generalising the work IBM has been doing around dynamic priority classes.

"The goal of this effort is to add support for the HyperShift Operator to ensure minimal disruption for customer control plane workloads that generate higher revenue.  Priority classes on the control plane workloads can be dynamically adjusted based on the number of worker nodes associated to the HC. 
This helps ensure that control plane pods of higher priority clusters don't get evicted/rescheduled before the pods of a lower priority cluster. Dynamically adjusting priority classes is something IBM Cloud has been doing to maintain and operate their Kubernetes/OpenShift service at scale."

https://github.com/openshift/enhancements/pull/1257