For larger etcd database sizes we should enable "experimental-stop-grpc-service-on-defrag". This allows the server to immediately reject requests instead of timing them out for a long period of time (it takes 5m to defrag 36 gigs).

This is very well described with: https://github.com/kubernetes/kubernetes/issues/93280

–

Waiting for https://github.com/etcd-io/etcd/pull/17914 to merge and to rebase with the next etcd 3.5 release - or we should carry it for the time being.