As a part of add / remove shard, we send the _shardsvrJoinDDLCoordinators command to all shards in the cluster with a read preference set to primary.

If any of these fail due to a FailedToSatisfyReadPreference exception, we will fail the overall add / remove shard command.

BF-33560 shows that if we have continuous stepdown running, we could get to a state where a replica set does not have an active primary and we receive this response when running the _shardsvrJoinDDLCoordinators command against the cluster.