Well, I've never had the resources to try this, but it might work.  Contract
for bucket loads of disk at your disaster site, enough to restore your
critical nodes' primary pool, and restore the copypool to the diskpool using
the maxpr switch to get multiple processes to speed the restore.
Noncritical nodes can wait on tapes.  Your critical nodes can then restore
quickly and simultaneously from disk.

Note, if anybody has the funds for that, please don't tell me as it would
make me too darned jealous.  Any thoughts on feasability, other than cost?

Alex