A new metric [1] was introduced in Kubernetes 1.23 for the kube-apiserver latency which doesn't take into account the time spent processing webhooks. This is meant to give a more accurate view of the SLO of the kube-apiserver and as such we should switch to it.

Note that this metrics was renamed in 1.26 [2] and the original one was deprecated and will be removed in 1.27.

[1] https://github.com/kubernetes/kubernetes/pull/105890
[2] https://github.com/kubernetes/kubernetes/pull/112679