(Note: All the UIs are giving me local time, currently UTC+10, so the breakage date is probably July 1st for readers located in less-advanced timezones. 😜)

Description

Every PR build after 1410704728173580288 has failed, and the ones I've looked at have failed the same way:

[sig-node] ResourceMetricsAPI [NodeFeature:ResourceMetrics]
_output/local/go/src/k8s.io/kubernetes/test/e2e_node/framework.go:23
  when querying /resource/metrics
  _output/local/go/src/k8s.io/kubernetes/test/e2e_node/resource_metrics_test.go:45
    should report resource usage through the resource metrics api [It]

A little digging shows it's not a containerd source change, since 5ccc77f65b58b7fe0df2120dbf9ab8c78b275b37 passed (prow) and c31caeea2b16e280f53f2769d3683c06618ba32e failed (prow) and these are both commits from #5610 (prow history) which differ trivially and unrelatedly, and notably haven't been rebased to the main branch between passing and failing.

There was a bump of the kubekins-e2e test image a few days before the problem appeared, but that was deployed within minutes of merging and there were subsequent successful builds which were presumably using that image.

Update: For visibility, the kubernetes-side bug for this failing test is kubernetes/kubernetes#103485