When running mapReduce in sharded environment, temporary collections are created within the source database, even if you specify a separate output database. When outputting to a collection in a separate database, it would be beneficial if those temporary collections could be created in that separate database.