Both problems stopped at about 20 minutes (I noticed the fan slowing down), while still in the 0% initial phase though (for new crunchers of HCMD WUs that means much before the first checkpoint).

Cheers. Jean.
Edit: Added the reference to the checkpoint.

One observation is, that the sciences do not listen for the throttle instruction during the initial set up. Obviously if the set up is a few minutes, no one notices, but with 20% it becomes audible (fan noise).

That was reported by one of the programmers in the past. As yet not seen it on my C2D, but certainly now 9:41 CPU hours and 71% all indicates normal in the TM..... they continue to run Veeeery Long.

To assure, though the science does run using all idle cycles, on pretty much most crunchers it has no effect. The program does step aside as always if you need the machine to do something else. Very large checkpoint saves may give an occasional hick due disk i/o.... with HCMD it's though that everything in RAM and virtual memory (VM) is small.