1. Proposed title of this feature request

Allow evicting unhealthy hosted control-plane pods

2. What is the nature and description of the request?

As WRKLDS-1490 points out, OpenShift 4.18 (Kubernetes 1.31) is GAing the PodDisruptionBudget unhealthyPodEvictionPolicy property. This request is asking the HostedControlPlane operator to populate that property for PodDisruptionBudgets that it manages.

3. Why does the customer need this? (List the business requirements here)

Setting that property to AlwaysAllow allows running-but-unhealthy pods to be evicted, which makes it easier to drain nodes on the management cluster, without much increased risk for the availability of hosted services (because the unready pods were not likely to be contributing to service availability).

Backing context on the new property.

4. List any affected packages or components.

WRKLDS-1490 is asking all PDB maintainers to think through this property, but this RFE is specifically about HyperShift's HostedControlPlane controller.

This will take some work, because the HostedControlPlane controller would have to use k8s.io/client-go/openapi3 or similar to detect whether the management cluster supports unhealthyPodEvictionPolicy. Without this detection, attempting to set the property on a management cluster that doesn't support it could result in the controller hot-looping on PDB update/patch calls.