*...

- Added a return value to `QueueingHint` to indicate an error. If `QueueingHint` returns an error,
  the scheduler logs it and treats the event as a `QueueAfterBackoff` so that
  the Pod won't be stuck in the unschedulable pod pool. ([#119290](https://github.com/kubernetes/kubernetes/pull/119290), [@carlory](https://github.com/carlory))
- Added apiserver identity to the following metrics: 

- Limit number of instances in a single update to GCE target pool to 1000. ([#87881](https://github.com/kubernetes/kubernetes/pull/87881), [@wojtek-t](https://github.com/wojtek-t)) [SIG Cloud Provider, Network and Scalability]

For stable features, the `kubectl` tool follows the Kubernetes [version skew policy](https://kubernetes.io/releases/version-skew-policy/);
however, kubectl v1.21 and older do not support the new API for ephemeral containers.

* Reduces GCE PD Node Attach Limits by 1 since the node boot disk is considered an attachable disk ([#80923](https://github.com/kubernetes/kubernetes/pull/80923), [@davidz627](https://github.com/davidz627))

- The scheduling queue didn't notice any extenders' failures, it could miss some cluster events,
  and it could end up Pods rejected by Extenders stuck in unschedulable pod pool in 5min in the worst-case scenario.

  - A namespaced object with an ownerReference referencing a uid of a namespaced kind which does not exist in the same namespace is now consistently treated...