c682d2b8c5
When filtering fails because a ResourceClass is missing, we can treat the pod as "unschedulable" as long as we then also register a cluster event that wakes up the pod. This is more efficient than periodically retrying.