TechPreview clusters are unable to bootstrap because kube-apiserver fails to start with the following error:

E0827 20:29:22.653501 1 run.go:72] "command failed" err="group version resource.k8s.io/v1alpha2 that has not been registered"

This happens because, in Kubernetes 1.31, the group version resource.k8s.io/v1alpha2 was removed and replaced with resource.k8s.io/v1alpha3. This is part of the DynamicResourceAllocation feature, which is currently TechPreview.

After discussing this with the team, we decided that the best approach is to modify the cluster-kube-apiserver-operator to start the kube-apiserver with the correct group version based on the Kubernetes version being used.