Daily quota for BRP4 (opencl-intel_gpu) tasks

I've just found an interesting log entry for my Intel i5 / HD 4600 host:

27/10/2013 17:25:59 | Einstein@Home | Sending scheduler request: To fetch work.
27/10/2013 17:25:59 | Einstein@Home | Reporting 1 completed tasks
27/10/2013 17:25:59 | Einstein@Home | Requesting new tasks for intel_gpu
27/10/2013 17:25:59 | Einstein@Home | [sched_op] CPU work request: 0.00 seconds; 0.00 devices
27/10/2013 17:25:59 | Einstein@Home | [sched_op] intel_gpu work request: 4.49 seconds; 0.00 devices
27/10/2013 17:26:02 | Einstein@Home | Scheduler request completed: got 0 new tasks
27/10/2013 17:26:02 | Einstein@Home | [sched_op] Server version 611
27/10/2013 17:26:02 | Einstein@Home | No work sent
27/10/2013 17:26:02 | Einstein@Home | No work is available for Binary Radio Pulsar Search (Arecibo)
27/10/2013 17:26:02 | Einstein@Home | (reached daily quota of 128 tasks)
27/10/2013 17:26:02 | Einstein@Home | Project has no jobs available
27/10/2013 17:26:02 | Einstein@Home | Project requested delay of 26560 seconds
27/10/2013 17:26:02 | Einstein@Home | [sched_op] handle_scheduler_reply(): got ack for task p2030.20130203.G175.61-00.23.C.b2s0g0.00000_3372_1
27/10/2013 17:26:02 | Einstein@Home | [sched_op] Deferring communication for 07:22:40
27/10/2013 17:26:02 | Einstein@Home | [sched_op] Reason: requested by project

I've not seen that before, but thinking about it, today is a 25-hour day (clocks went back an hour in Europe this morning), and with even an i5 completing tasks in under 11 minutes, I may have been on the brink of exceeding quota on other days too.

Two questions arise.

1) When do quota days begin and end? The 7-hour backoff suggests 'midnight, plus a random fraction of an hour to stop everyone asking at once'. But when I finally noticed that the co-processor had been idle for over three hours, I was able to get new work without complaint:

27/10/2013 17:54:31 | Einstein@Home | Computation for task p2030.20130203.G175.61-00.23.C.b4s0g0.00000_2948_1 finished
27/10/2013 21:14:11 | Einstein@Home | update requested by user
27/10/2013 21:14:12 | Einstein@Home | Sending scheduler request: Requested by user.
27/10/2013 21:14:12 | Einstein@Home | Reporting 3 completed tasks
27/10/2013 21:14:12 | Einstein@Home | Requesting new tasks for intel_gpu
27/10/2013 21:14:14 | Einstein@Home | Scheduler request completed: got 3 new tasks

2) Would it be a good idea to think about the packaging, and labelling, of this application? There have been some comments recently by people thinking 'My core-I processor has an internal GPU, I'll select the GPU application for it', and finding that they don't get allocated the 16-pack tasks intended for large, discrete, GPUs. On the other hand, maybe the single packs designed for CPUs are too small for this type of processor.

Goldilocks would like a 4-pack or 6-pack workunit, flagged specifically for iGPU...