rman allocate channel suddenly taking forever

Hi all,  

We run an rman level 0 backup of a Dataguard physical target database, twice a week, to disk. The first thing the job does is allocate 10 disk channels. This part of the job (the allocate commands) usually takes 40 minutes to a little over an hour. For the last 3 runs of this backup job, however, the allocate is taking over 9 hours.  

I should note that this is a very large database with 10's of thousands of data files. This is Oracle 9.2.0.7 64-bit on Solaris.  

Any suggestions as to what we might look at? What exactly happens internally during the allocate channel command? My rman catalog db is generating 3.1 Gb or archive logs during the 9 hours that the allocate channel commands are running.  

Thanks for any help or pointers... we've searched Metalink but came up with nothing.    

Jerry

--
http://www.freelists.org/webpage/oracle-l