What is the hashrate that's reported by your miners?
I uncovered another little flaw in the hashrate tracking algorithm, that I will fix now, but that doesn't explain much of a deviation.
However, while looking at the logs, I saw your miners (you're PS2KDRDhDt2X94yPrVnTe2rqLitunL84XF, aren't you?) behaving really weird and sending shares with garbage job IDs, as if they would suffer from some kind of memory corruption. You'll probably want to restart them. Did they report a large amount of share upload failures recently?