* **beta** PodSecurityPolicy objects limits use of security-sensitive features by pods.
* Kubectl
  * Display line number on JSON errors
  * Add flag -t as shorthand for --tty
* Resources
  * Improved node stability by *optionally* evicting pods upon memory pressure - [Design Doc](https://github.com/kubernetes/kubernetes/blob/release-1.3/docs/proposals/kubelet-eviction.md)

### Bug or Regression

- Fixed a bug where a high churn of events was causing master instability by reducing the maximum number of objects (events) attached to a single etcd lease. ([#100452](https://github.com/kubernetes/kubernetes/pull/100452), [@mborsz](https://github.com/mborsz)) [SIG API Machinery and Instrumentation]

### Apps

The core workloads API, which is composed of the DaemonSet, Deployment, ReplicaSet, and StatefulSet kinds, has been promoted to GA stability in the apps/v1 group version. As such, the apps/v1beta2 group version is deprecated, and all new code should use the kinds in the apps/v1 group version.

### Auth

- Kube-apiserver: fixes a 1.27+ regression in watch stability by serving watch requests without a resourceVersion from the watch cache by default, as in <1.27 (disabling the change in #115096 by default). This mitigates the impact of an etcd watch bug (https://github.com/etcd-io/etcd/pull/17555). If the 1.27 change in #115096 to serve these requests from underlying storage is still desired despite the impact on watch stability, it can be re-enabled with a `WatchFromStorageWithoutResourceVersion`...

available only if you\n   * include `popper-utils.js` before `popper.js`.\n   *\n   * **DEPRECATION**: This way to access PopperUtils is deprecated\n   * and will be removed in v2! Use the PopperUtils module directly instead.\n   * Due to the high instability of the methods contained in Utils, we can't\n   * guarantee them to follow semver. Use them at your own risk!\n   * @static\n   * @private\n   * @type {Object}\n   * @deprecated since version 1.8\n   * @member Utils\n   * @memberof Popper\n   */\n...

    NVIDIA Ampere based GPUs and is enabled by default.

*   A major refactoring of the internals of the Keras Functional API has been
    completed, that should improve the reliability, stability, and performance
    of constructing Functional models.

*   Keras mixed precision API
    [`tf.keras.mixed_precision`](https://www.tensorflow.org/api_docs/python/tf/keras/mixed_precision?version=nightly)

## External Dependencies

   ([#113329](https://github.com/kubernetes/kubernetes/pull/113329), [@RomanBednar](https://github.com/RomanBednar))
- '`registered_metric_total` will now report the number of metrics broken down by
  stability level and deprecated version.'
   ([#112907](https://github.com/kubernetes/kubernetes/pull/112907), [@logicalhan](https://github.com/logicalhan))

tolerate the NoSchedule taint if the taint is added to a Node.  There is an open PR ([#48189](https://github.com/kubernetes/kubernetes/pull/48189)) to resolve this issue, but as this issue also exists in 1.6, and as we do not wish to risk release stability by merging it directly prior to a release without sufficient testing, we have decided to defer merging the PR until the next point release for each minor version ([#48190](https://github.com/kubernetes/kubernetes/pull/48190)).

* Protobuf...

- Kube-apiserver: fixes a 1.27+ regression in watch stability by serving watch requests without a resourceVersion from the watch cache by default, as in <1.27 (disabling the change in #115096 by default). This mitigates the impact of an etcd watch bug (https://github.com/etcd-io/etcd/pull/17555). If the 1.27 change in #115096 to serve these requests from underlying storage is still desired despite the impact on watch stability, it can be re-enabled with a `WatchFromStorageWithoutResourceVersion`...