Router pods scheduled to same node on HCP with SingleReplica availability policies leading to eviction loop

Issue

• In a Hosted Control Plane (HCP) cluster where infrastructureAvailabilityPolicy set to SingleReplica, ingress router pods are repeatedly scheduled onto the same node.
• The Ingress Operator detects this pod placement as invalid (expecting router pods to be distributed across nodes) and continuously evicts the router pods.
• This results in an endless eviction and rescheduling loop, preventing ingress from stabilizing and causing degraded or unavailable ingress functionality.