data = response.json()
    assert data["context_b"] == "started b"
    assert data["context_a"] == "started a"
    assert data["sync_bg"] == "not set"
    assert state["context_b"] == "finished b with a: started a"
    assert state["context_a"] == "finished a"

- Scheduler now waits for handlers to finish syncing before the scheduling cycles start. ([#116729](https://github.com/kubernetes/kubernetes/pull/116729), [@AxeZhan](https://github.com/AxeZhan))

- KCCM: fixed transient node addition and removal caused by #121090 while syncing load balancers on large clusters with a lot of churn. ([#121091](https://github.com/kubernetes/kubernetes/pull/121091), [@alexanderConstantinescu](https://github.com/alexanderConstantinescu))

- A new feature was enabled to improve the performance of the iptables mode of `kube-proxy` in large clusters. No action was required, however:
  

- Increased the maximum backoff delay of the endpointslice controller to match the expected sequence of delays when syncing Services. ([#112353](https://github.com/kubernetes/kubernetes/pull/112353), [@dgrisonnet](https://github.com/dgrisonnet))

- Fixed performance regression in JSON logging caused by syncing stdout every time error was logged. ([#107035](https://github.com/kubernetes/kubernetes/pull/107035), [@serathius](https://github.com/serathius))

* [fluentd-gcp addon] Fluentd will trim lines exceeding 100KB instead of dropping them. ([#52289](https://github.com/kubernetes/kubernetes/pull/52289), [@crassirostris](https://github.com/crassirostris))
* dockershim: check the error when syncing the checkpoint. ([#52125](https://github.com/kubernetes/kubernetes/pull/52125), [@yujuhong](https://github.com/yujuhong))