The crashes with bad alloc all come from the same set of runs, which share in common a different type of input file. I'm not sure what's causing the crash so in the meantime, I've eliminated these runs from the queue. Work units that have already been sent out will still run (and it looks like the crashes are rare events) but they should be ending within the next day or two. I'll let you know before I relaunch this type of job what was the problem. Thanks, Sarel.