Bisect points to https://chromium-review.googlesource.com/c/chromium/src/+/1384137 as the culprit. It looks like the tests were failing on the try-nougat-phone-tester runs on that patch, as well, but weren't visible due to the builder failing to retrieve the results from swarming.

eirage@ did some initial debugging, and it looks like we are flakily actually sending scroll events out of order in VR, so I'll take this back.

Fix is available at https://chromium-review.googlesource.com/c/chromium/src/+/1394170, although the one non-test VR Java OWNER is OOO until next week.

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/e454a7b40f714c0bca5851bbd07aaad2b7fc2fb2

commit e454a7b40f714c0bca5851bbd07aaad2b7fc2fb2
Author: bsheedy <bsheedy@chromium.org>
Date: Mon Jan 07 18:35:00 2019

Fix out of order VR scroll events

Fixes scroll events in VR flakily being sent out of order, which was
causing a DCHECK to be hit, and less frequently, for the scroll tests
to flake due to not scrolling the expected amount.

This is done by dropping any delayed events that are delayed for long
enough for another event with a newer timestamp to be injected before
the delayed one.

Bug:  918580 
Change-Id: Id3b170455b052dc51193c9eb65ce1093fd593f90
Reviewed-on: https://chromium-review.googlesource.com/c/1394170
Reviewed-by: Michael Thiessen <mthiesse@chromium.org>
Commit-Queue: Brian Sheedy <bsheedy@chromium.org>
Cr-Commit-Position: refs/heads/master@{#620391}
[modify] https://crrev.com/e454a7b40f714c0bca5851bbd07aaad2b7fc2fb2/chrome/android/java/src/org/chromium/chrome/browser/vr/AndroidUiGestureTarget.java