Some metadata commands have different behavior on retry (for example, re-running dropCollection on a standalone causes the second attempt to return an error, whereas on a sharded cluster, the second attempt returns success) when run on a sharded cluster vs. on a standalone. We should standardize behavior across the two.