Not sure this will be relevant but your system of 12 cards using one 850 psu for the risers is how I killed two psu. I only had 7 cards running and it was too much for the Sata channels to handle. One psu will be good for 6 sata devices so I think you are browning out the main psu. Once I dropped back on card count, no more problems. In the end, I gave up on the Server psu's and went with dual 1000w psu's.
thay
That is a good point, Thay. I was very surprised to learn how much some GPUs pull through the PCI slot and also that it varies by GPU and MFR. If we assume 50w per card (may even be low) and another 80w for the mobo, that gives us 680w. That is what I would consider to be the max usable for a 850w PSU (80% of rated output).
EDIT: IMO, trying to get a 12 GPU rig to be rock stable is an exercise in frustration. I have read dozens of messages on this thread (and others) of folks having all kinds of oddball problems with rigs that use more than 8 GPUs. This includes not only nvOC, but also Windows, SMOS, and HiveOS. It can be done, but it comes with an inherent lack of stability. If I were you, I would split it up into 2 rigs and get on with mining.