Strange that RMAN doesn't differentiate a database (not really a new 'incarnation') restored under the below circumstance. (And yes, I realize that only a "resetlogs" leads to a new incarnation). Wonder how RMAN will go about restoring/recovering in the below scenario. It might be worth a test sometime! Hopefully, I'm not missing something obvious..
Say, Database Db1 is in archivelog mode.
1) It was shutdown a week ago. A cold backup (with a simple file copy and NOT using RMAN) was taken (including d/fs, redologs, ctl files). The database was then at sequence # 100.
2) A day later you start to use RMAN on Db1, with a separate catalog - RMANCAT. You're happily crusing along backing up archivelogs and the like. The sequence#s are incrementing in the meanwhile.
3) A few days later, you're at sequnce# 200. Now, you wipe out DB1.
4) Restore Db1 from step 1 (cold backup) with a simple "startup". You're back to sequence# 100.
5) Now start backing up the database, archivelogs etc using RMAN again. Note : you didn't have to create a new incarnation!
6) A few days later, try restores and recovery (with SCNs) on this new database. I would think it ought to confuse the hell out of RMAN!
(Worse yet, in step 5 - say your RMAN script issues a 'backup archivelog all not backed up 1 times;' - the script will probably skip many archivelog files altogether, since it would incorrectly think that they'd already been backed up!)