Backup Architecture question

OK, this one's pretty ugly. Let me start with some background.

Oracle 9.2.0.5 server. One data instance, one repository instance. Database is up 24/7, with hot backups nightly run from OMS. Nightly, the OMS/agent/repository is shut down automatically, and a tape backup of the /opt partition is made. If the disk fries, we can restore the partition from the backups, start the repository instance (& agent & OMS), then restore the main database from the hot backups.

That all works, we think. (System is still in development, but passed the surface testing I did.)

Ah, but they have a new, cunning plan. In the event of an Oracle patch, they want to ship out a new hard drive with the patched $ORACLE_HOME, then restore the database from the backups. Of course, all they want restored from the backup is the actual database, not anything that might have been superceded by the patch.

And we now hit the wall of my Oracle knowledge. The reason for the cold backup of the repository is that I couldn't get it to run after a restore taken while it was running, and I didn't know how to bootstrap  it.

What is the set of files that have to be restored from tape in order to permit the reconstruction of the repository database w/o impacting any files that may have been patched? (If I can get the repository up & running, I believe all I need is the most recent backup set made by the OMS to reconstitute the data instance and I'm back in business, right?)

Or is that simply such a Wrong Way To Do It that I have to present them with a Plan B?

-- 
Jeremy Billones