Why not put them all in the same dir if you can manage. The archive format will show you the thread so you'll know where each log came from and RMAN only has to backup the one dir which is sort of kitchy.
Also, if you writing to a SAN or shared file system try and duplex
the logs to write their and the local file system as well if possible. Elimates the shared storage device as a single point of failure.
Actually we already have a 4 node RAC cluster implemented for over a year now. We are writing the archive logs to 1 device. But what we have noticed recently is that each of the nodes may do a log switch at the same time, and they may all write 500M to the archive file (that is on 1 device) at the same time.
So, if we can have a dedicated device for each instance, then that should make the archieving faster; without waiting for the device to free up.