I have been searching signs of mechanical hit but not found any evidence. Also there is a bit flex in the machine because of the dimensions and it not being made out of cast iron so in these speeds and with these forces the conical nozzle should be able to climb a bit if meeting a blob. I have been manually testing how much force there has to be applied to make the X and Y skip steps and it is considerable.
The current is around 2 amps, I have been playing on both sides of that in search of well running settings with my external DM542A and DQ542MA drivers. The motors are separated from the build chamber, enclosed and cooled. Drivers stay nice and cool.
I was earlier not able to get the Zhop working (using Cura) but now will put more effort on that. It may have been caused by the short signal width that I now have changed into 3 microseconds.
The pulleys and belts and shafts are tight and playfree.
Temperature related binding in the linear rail has been under consideration but they are from a quality manufacturer and I would expect the error then continuing from a certain timepoint on or at least be more frequent. Instead the layer shift happens only every now and then, and after a shift the print continues beautifully untill the next shift that comes randomly at some point.
Power related issues have been thought too, but they're not probable with the UPS (and surely would be weird affecting only X as you said.)