If I happen to request work for a CPU (which you don't have), I get a very misleading error message from the server:

03/06/2010 10:29:28 GPUGRID [sched_op_debug] Starting scheduler request
03/06/2010 10:29:28 GPUGRID Sending scheduler request: To fetch work.
03/06/2010 10:29:28 GPUGRID Requesting new tasks for CPU
03/06/2010 10:29:28 GPUGRID [sched_op_debug] CPU work request: 1.00 seconds; 0.44 idle CPUs
03/06/2010 10:29:28 GPUGRID [sched_op_debug] NVIDIA GPU work request: 0.00 seconds; 0.00 idle GPUs
03/06/2010 10:29:33 GPUGRID Scheduler request completed: got 0 new tasks
03/06/2010 10:29:33 GPUGRID [sched_op_debug] Server version 611
03/06/2010 10:29:33 GPUGRID Message from server: No work sent
03/06/2010 10:29:33 GPUGRID Message from server: Fermi-class GPU needed
03/06/2010 10:29:33 GPUGRID Project requested delay of 31 seconds

It would be better if the message conveyed, more accurately, "We don't use CPUs for crunching on this project."

But - correct me if I'm wrong - you do still supply work for the 9800GT that's in this box, but busy on other projects?