I am old enough to have started in "Data Processing," where we handled 3,000-card trays of punched cards stacked 6 feet high on carts. We punched them, sorted them, "proved" them and then ran them through an IBM 407 (which required days of manual wiring) to get a printout.
The reason this may apply to this current problem is that we would get through a certain stage of processing, go home for the night, and come back in the morning to find the cards in an unexpected order, other than the one in which we left them.
We finally named the situation "osmiotic unsort," discovering that the little holes in the punched cards ran around at night and didn't always get back to where they came from before we arrived the next morning.
Maybe a similar grinch is working here on the electronic level?
By the way, we never did catch the little holes in motion and I imagine that electrons move much faster, so good luck.
Larry