RMAN is doing the same trick at the DF level. when RMAN do the copy file, it's like hot backup.
actually, in metalink, there is an internal note about how to use RMAN backupset if the control file is corrupted and unrecoverable. So essentially, controlfile is not needed, all needed is the extracted datafiles from the backupset.
In my experiment, one more complication is the RAC ASM -> non-RAC/non-ASM. The data files are all magic numbers generated by Oracle Managed files. When I do the DB refresh, I will stop the Standby DB apply (non-RAC/non-ASM). and copy the DFs over. This provides consistency during the copy.
Havent done this, but we explored the possiblity of doing this for a cross platform move from Sun to IBM. You can achieve this for one-time or repeated full refresh by using RMAN to move from one platform to another platform in 10g (depending on the "endian" format of the db). Your case is not a one-time refresh. You have a different requirement. Your requirement is to keep the data in a non-production environment in sync with production. You might be able to use STREAMS. I am not sure about it cos of cross-platform issues. OR you can use plain-old replication.