The problem could be reproduced on openshift and 4-node k3s cluster.

Steps to reproduce:
setup a fresh cluster and install operator:

Setup minio with helm, merge backup configuration to cr.yaml and apply cr.yaml

Apply backups one by one (change backup name) until psmdb status became error:

If a cluster in the error state backup is failed.

operator produces periodic messages with new error:

During error state mongo clients connecting slowly (ssl enabled):

Number of connections slowly grows:

the mongo server is reachable from operator host:
curl -k https://my-cluster-name-rs0.default.svc.cluster.local:27017/
curl: (52) NSS: client certificate not found (nickname not specified)

operator has many tcp connections opened:

after kubectl delete pod percona-server-mongodb-operator-588db759d-fcgww
psmdb status returns to normal and backup are possible again.

The stale could be related to similar issue at:
https://jira.percona.com/browse/K8SPSMDB-271

The error is consistently could be reproduced on my host (if it's idle, if CPU is used by other tasks and kubernetes is slow backup is not causing psmdb error and further backup errors)