See if you can check which process is plotzing. I'm not at my machine right now but if I remember correctly, fah.exe starts mpiexec.exe; and smpd.exe calls the four core instances. Run a process explorer of some kind and see if you can tell which one fails, and what goes down with it.
What exactly does the log say? Does it just stall at a % and go no further?
Is the notebook freestanding or is it on your network when it happens?
I'll assume that no overclocking is going on. Or is there? If you are pushing the FSB then back off a touch and see if the problem persists. Even if the machine has turned out dozens of WUs in the past, new generations can stress the machine in different ways. If you're overclocking I'd put my money on that being the cause. Try going to stock and run the same WU again, even if it's expired. [You are running a backup, aren't you?
] Or get another one from the same project and run it at stock to see.
I'm still getting 2653's on all 40 boxen. I think there's a new project out there that I haven't seen as yet. Did you land one of those by any chance?
There's one more possibility. How long did you wait before terminating the window and re-starting? In the FAHWiki I remember reading that on some WUs the speed can increase or slow down dramatically at some points, for varying amounts of time. This is called an "event" and is related to some dynamics in the protein folding process; it's mathematical and not related to your hardware. I've not actually witnessed one, but I guess if a user were "lucky" enough to be at the screen when one occurred, it would appear that the WU had stalled [or suddenly went into high gear.] Doesn't sound like that is the case here but it's worth considering.