Right...probably the other one :)

It is not related to  issue 715742 , since the cc animation is correct.
It is not related to  issue 756740 , since that patch cause  issue 756740  reverted.

It looks like a new hit test issue.

Ah, right, I got confused as to what that was tracking and assumed this might be squashing but since the CC animation happens I guess it can't be squashing and must be related to hit testing.

This issue is caused by the size of overlay scrollbar is incorrect, in my debug output, the scrollbar rect(h_bar_rect)[1] is thin rect but scrollbar already expand I check the FrameRect is correct. You can move the mouse to almost the bottom of scrollbar then you can drag the scrollbar.

Tried change the Y of scrollbar calculation to: (Box().BorderTop().ToInt() + VisibleContentRect(kIncludeScrollbars).Height() - HorizontalScrollbar()->ScrollbarThickness()). Then the hit test include scrollbar.

Then I found another issue for this: ScrollbarTheme::HitTest return NoPart. ConvertFromRootFrame return an incorrect point, it seems something calc twice. Need more investigation.

Demo: http://ht.chaopeng.me/scrollbar-hit-test-wrong.html

[1] https://cs.chromium.org/chromium/src/third_party/WebKit/Source/core/paint/PaintLayerScrollableArea.cpp?rcl=a16189bedba98586e1360666460717f88b63d8ee&l=1518

This is a regression introduced from 01cf6fd553bcc6d5a9ca6763eed42fbca5ecc494. 
Stefan, Could you please take a look. Thank you.

ping? This is P1 but no action in a while.

I'm unable to reproduce this in version 63.0.3239.9 (Linux desktop).  I can scroll the region using shift-wheel.

The overflowing element has overflow:auto, so I'm not sure why comment #5 mentions overlay scrollbars.  I'm also unable to reproduce the demo from that comment.

szager@ this needs turn on overlay scrollbar

--enable-features=OverlayScrollbar --enable-prefer-compositing-to-lcd-text

I can reproduce at 64.0.3258.0

Can not reproduce in CrOS in 67.0.3833.0 and Linux in ToT.