assertContent("This connection won't pool properly", response1)
    assertThat(server.takeRequest().exchangeIndex).isEqualTo(0)

    // Give the server time to enact the socket policy if it's one that could happen after the
    // client has received the response.
    Thread.sleep(500)
    val response2 = getResponse(newRequest("/b"))
    response1.body
      .source()
      .timeout()

- Overall, enable the feature of `PreferNominatedNode` will  improve the performance of scheduling where preemption might frequently happen, but in theory, enable the feature of `PreferNominatedNode`, the pod might not be scheduled to the best candidate node in the cluster. ([#93179](https://github.com/kubernetes/kubernetes/pull/93179), [@chendave](https://github.com/chendave)) [SIG Scheduling...

- Fix a bug that when PodTopologySpread rejects Pods, they may be stuck in Pending state for 5 min in a worst case scenario.
  The same problem could happen with custom plugins which have Pod/Add or Pod/Update in EventsToRegister,

This release continues to enable more existing features on Windows, including container CPU resources, image filesystem stats, and flexvolumes. It also adds Windows service control manager support and experimental support for Hyper-V isolation of single-container pods.

### OpenStack

- DRA: The scheduler plugin prevented abnormal filter runtimes by timing out after 10 seconds. This was configurable...

                # versions (as of writing 0.117.1) that allowed doing
                # except HTTPException inside a dependency with yield.
                # This needs to happen after user middlewares because those create a
                # new contextvars context copy by using a new AnyIO task group.
                # This AsyncExitStack preserves the context for contextvars, not

- Scheduler bugfix: Scheduler doesn't lose pod information when nodes are quickly recreated. This could happen when nodes are restarted or quickly recreated reusing a nodename. ([#93938](https://github.com/kubernetes/kubernetes/pull/93938), [@alculquicondor](https://github.com/alculquicondor)) [SIG Scalability, Scheduling and Testing]

  * Note that NumPy's type promotion rules have been changed(See [NEP 50](https://numpy.org/neps/nep-0050-scalar-promotion.html#nep50)for details). This may change the precision at which computations happen, leading either to type errors or to numerical changes to results.
  * Tensorflow will continue to support NumPy 1.26 until 2025, aligning with community standard deprecation timeline [here](https://scientific-python.org/specs/spec-0000/).

</p>

<p>
For example, in the (function-local) assignment
</p>
<pre>
y[f()], ok = g(z || h(), i()+x[j()], &lt;-c), k()
</pre>
<p>
the function calls and communication happen in the order
<code>f()</code>, <code>h()</code> (if <code>z</code>
evaluates to false), <code>i()</code>, <code>j()</code>,
<code>&lt;-c</code>, <code>g()</code>, and <code>k()</code>.

- Fixed an EndpointSlice Controller hashing bug that could cause EndpointSlices to incorrectly handle Pods with duplicate IP addresses. For example this could happen when a new Pod reused an IP that was also assigned to a Pod in a completed state. ([#115907](https://github.com/kubernetes/kubernetes/pull/115907), [@qinqon](https://github.com/qinqon)) [SIG Apps and Network]