Same thing happens to me. Seems to happen more frequently with scrypt-n than scrypt for me, but I have no data to back that up. I *think* it may have to do with waking the monitor, it never happens on my machine that has monitor sleeping disabled, but the machine that does put the monitor to sleep locks up occasionally. Once I checked the pool it was running against and it had been submitting work right up until the moment I tried to use the machine (thus waking the monitor). It's a weak hypothesis, but who knows...