PrometheusAdapter and MetricsServer tasks are conflict prone

The two tasks will always "UPDATE" the underlying "APIService" resource even when no changes are to be made.
This behavior significantly elevates the likelihood of encountering conflicts, especially during upgrades, with other controllers that are concurrently monitoring the same resources (CA controller e.g.). Moreover, this consumes resources for unnecessary work.

The tasks rely on CreateOrUpdateAPIService: https://github.com/openshift/cluster-monitoring-operator/blob/5e394dd9de305cb6927a23c31b3f9651aa806fb8/pkg/client/client.go#L1725-L1745 which always "UPDATE" the APIService resource.

keep a cluster running then take a look at audit logs concerning the APIService v1beta1.metrics.k8s.io, you could use: oc adm must-gather -- /usr/bin/gather_audit_logs

    1.
    2.
    3.
    

You would see that every "get" is followed by an "update"

You can also take a look at the code taking care of that: https://github.com/openshift/cluster-monitoring-operator/blob/5e394dd9de305cb6927a23c31b3f9651aa806fb8/pkg/client/client.go#L1725-L1745

"Updates" should be avoided if no changes are to be made.

it'd be even better if we could avoid the "get"s, but that would be another subject to discuss.