I just wanted to let you know:
Of the 16 tasks I had since Feb 1 until an hour ago, only the second one (already on the first day, Feb 1) erred out and I can´t repro this any more.

But the point is:
When I just (Feb 19) resumed GPUgrid I had run the FaH GPU client before and the next two GPUgrid tasks failed:
323401 Zj10716-SH2_US_6-4-10-SH2_US_61750000_0
323388 dz12780-JAN4-17-25-acemd_0

I restarted the machine and the next task is just running fine!
Cheers
G.