They didn't all die at once, rather one at a time over a period of months. A few of them were already hashing at the much reduced rate of 30TH rather than their nominal 45TH, so they were already running at relatively tame current levels. I'll open up and do a close inspection of a few of those PSUs; I'm sure the hashing boards themselves are all fine, so I'm suspecting either the PSUs or controller boards at this point.