I saved this one at the point at the end where the error is ABOUT to occur, as soon as you copy Region 4 to it's end to produce Region 5. Seemed like a good idea, but maybe the problem is introduced even earlier. Dunno.
It appears the problem is at the joint. (If I jiggle the segment or last point there before copying, it "pastes" an apparently identical/correct region. Not shown.)
I show a "paste" starting at bar 13 that apparently succeeds, then undo that [0:05-0:08],
Then "paste" at the end of Region 4 itself [0:10], This "paste" fails to produce a region with length of whole number of measures,
Then undo that to prepare to save[0:12].
I just zoom in a little at the end so you can see that I have lines like heda in the decimal places. Bonus. (But still no way to tell from marker lanes alone, that there are 2 points at the same position.)
PS, it seems a change occurred that made the errors occurring in pre 15, which seemed to be at the start of the region, go away, at least in the limited testing I have done so far. In those ones, the points within the region shifted as well.
Now the error appears to be at the end segment. Points remaining on grid otherwise in the region that has the error.