Hello,
besides that the log cannot be backed up by TSM you are on the right way, I think. We have done similar things before with TSM 5.1.8 on z/OS. We had to change a RVA against an ESS.
Our TSM server runs on a 2064-102 with REGION=512M. The catridges are provided through a VTS/ATL combination by DFSMSrmm.
We had about 20 x 3390-3 mirrored DB volumes and 2 x 3390-9 mirrored recovery log volumes (mirrored as I mention is two copies from TSM's point of view, not PPRC or FlashCopy).
To move one copy of the mirrors we established the third copy on the new ESS location and afterwards deleted the copies on that old RVA volumes.
Of course first we did a DB backup. Actually we did two backups, one standard type=full and one type=dbsnapshot. Having a dbsnapshot in the backhand was good, because while defining, synchronizing and deleting among the mirrors obviously changed the database and the db backuptrigger started incremental copies a few times. Returning to the base situation may had become a struggle with all these full and incremental backups.
After all it took us three days (we changed one volume after the other, not all at the same time) and we had no trouble. The movement was done while TSM operation continued (primary goal) and the effect to performance was not that heavy as we expected.
Good luck, cheers
Michael