All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=783698

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=f0557c81e38c0ad4157983a4e56b1f0ae64c2adaabb888f690d28daea0a411be


Bot(s) for this bug's original alert(s):

chromium-rel-win7-gpu-nvidia

This is weird. The bisect didn't kick off because there is only one CL in the range:

r512419
> [root layer scrolls] Rebaseline 11 passing tests

> This patch rebaselines 11 passing tests. These baselines now
> match the non-RLS path. The updated expectations were generated by
> copying the non-RLS expectation and adjusting for the scrolling
> layer differences.

I looked at the source files, it does indeed only touch layout tests. Strangely, the ref build improves a few revisions later. Not sure what's going on here, but don't think we'll get to the bottom of it :(