From the database point of view there are no duplicate ROWIDs. If there were the database would be corrupt. I am just guessing the query is used to populate TEMP-TABLES and these TEMP-TABLES do have unique indexes defined on a field where either the ROWID of the database record or a TEMP-TABLE record is stored and then, caused by the OUTER-JOIN in your query, the same record occurs more than once in the result of your query.
Again, that's is just a guess!
Heavy Regards, RealHeavyDude.