SERVER-132175 added a test case in check_metadata_consistency.js that makes a few snapshot reads fail with SnapshotUnavailable and then checks that checkMetadataConsistency retried internally.

However the failSnapshotReads failpoint is not selective for checkMetadataConsistency, so the retries may be consumed by background activity (e.g. catalog cache, ShardRegistry, ClusterServerParameterRefresher), so the check that the error was retried internally by checkMetadataConsistency will fail.

We should make the failpoint specifically target checkMetadataConsistency, excluding background activity.