For counters an internal cache is used ___counter_configuration which is started eager.
If counters are not used and there is a cluster split (only two nodes) the cache would be degraded and a new node (i.e. one pod which is started to replace the crashed node) is not able to join the cluster.
In this case the livenessprobe will fail for OpenShift.
There is no way to change the PartitionHandling for this internal caches.

If the counters are not used the cache should be inactive to not consume resources or cause issues.