As discussed in this CL:

https://chromium-review.googlesource.com/c/chromium/src/+/752030

PaintLayer::PixelSnappedSize() uses Box().Location() to do the snapping, but Box().Location() is not the same as paint offset.

There are four places where layer size comes into play:

1) Determining overflow -- PaintLayerScrollableArea::Has(Horizontal|Vertical)Overflow

2) Determining clip rects -- PaintLayerClipper::CaluclateRects

3) Positioning overflow controls for paint -- PaintLayerScrollableArea::PositionOverflowControls

4) Hit testing -- PaintLayerScrollableArea::HitTestOverflowControls

Currently, all of these use a layer size snapped to box.Location(), but they *should* all use the accumulated paint offset for snapping.  That offset is available to the clipping code (2), and could be plumbed into the overflow control positioning code (3) fairly easily.  Passing it to overflow computation (1) and hit testing (4) will be harder.

The worst thing that can happen with the current code is that layer size will be off by one pixel; so not the end of the world, but still not correct.