I can tell you that to some extent I am already doing what you described.  I
have two STK Powder Libraries with 10 9840FC drives between the two
libraries that are located in different building (about 2 miles apart), a
Win2000 Cluster in one location and the disk for that cluster in the second
location.  Everything is connected by fiber switches and the 2 fiber
switches are connected through dark fiber.  The database backup is performed
to the library were the physical server nodes are so that it is away from
the disk that contains the live database.  The Library that is with the
servers contains the tapepool and the second library were the disk are
contains the copypool.  Disaster recovery is kind of like this.

If building with physical servers, primary tapepool and database backup is
destroyed then I would build a server in the backup location were the disk
is located (database is already there in production) and then mark the
primary tapes as destroyed.  Should be up and running in the amount of time
that it take me to build a Win2000 machine for the most part.

If building with the disks are destroyed then I would have to get new disk
connected to SAN and then restore the database and delete are copypool tapes
and recreate them.  The Win2000 server is already built so the recover would
be the time it takes to restore the database.