To run in suites with forced failovers, the add remove shard background hook will retry the commands it runs on expected retryable errors, e.g. here for the balancerStop command. This is based on a list of errors gathered from the driver spec, which does not include FailedToSatisfyReadPreference. For commands that run on mongos against the config server, if the config server is frequently failed over, they can fail with that error, so to avoid failures, we should make the hook retry on FailedToSatisfyReadPreference.