This isn't a new problem, as it turns out. The double difference has been bad for a while... how did we miss this?
Going back in time to look at a few randomly selected runs (all 960Hz, all 140-150 uA).
Run 7534 - 11/24 (PC off, from today)
Run 7409 - 11/21 18:29 DD~197 ppm (the last high current run before yesterday)
Run 7389 - 11/21 03:49 DD = 453
Run 7364 11/20 07:18 DD= 547.5
Run 7286 - 11/19 00:32 DD= 718 (pattern#<70k, to avoid beam trip) Run 7239 - 11/17 DD = 152.