Okay I did a manual bisect and this is the regression range:

https://chromium.googlesource.com/chromium/src/+log/5cb0339fb2376e0096f41459e46ce41b51d66ac4..900dd0f1b5215ce5146c08f71b21894065f7efe7

sdy@: Any idea which CL in this range could have cause this issue?

Thanks in advance.

900dd0f1b5215ce5146c08f71b21894065f7efe7 definitely looks the most plausible. I see viz show up in stack traces a bunch. fsamuel@, what do you think?

You said you see viz show up in stack traces. Do you have any sample stack traces?

I'm definitely able to repro this locally on 67.0.3370.0. I am also able to repro this tip fo tree. I've tried with and without surface synchronization. Repros..

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/cfb855831a545f535565625209fe437017e05525

commit cfb855831a545f535565625209fe437017e05525
Author: Fady Samuel <fsamuel@chromium.org>
Date: Thu Mar 15 23:30:17 2018

Display Scheduler: Don't request BeginFrames while waiting for a root surface

When the root surface changes, we cannot draw to the display until a surface
activates. This CL stops draws (and BeginFrames) until activation. If a root surface is never
submitted then the display will not listen for BeginFrames.

Bug:  821940 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel
Change-Id: I749c0ca8e85a3dbc7eccaa0799add81afb15beb9
Reviewed-on: https://chromium-review.googlesource.com/964595
Reviewed-by: Eric Seckler <eseckler@chromium.org>
Commit-Queue: Fady Samuel <fsamuel@chromium.org>
Cr-Commit-Position: refs/heads/master@{#543548}
[modify] https://crrev.com/cfb855831a545f535565625209fe437017e05525/components/viz/service/display/display_scheduler.cc
[modify] https://crrev.com/cfb855831a545f535565625209fe437017e05525/components/viz/service/display/display_scheduler_unittest.cc

I'd like to merge this back into M66 by around Monday in time for the next M66 beta build.

Have you confirmed the fix in Canary?

For me, from the reporter-view, this looks fixed in latest Canary Version 67.0.3372.0 on macOS 10.12.6.

Thank you fsamuel@ for the quick fix!!!

Approving merge for 66. Branch:3359

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/8adbb60e9662cc609447c691137aa88c38f71b8f

commit 8adbb60e9662cc609447c691137aa88c38f71b8f
Author: Fady Samuel <fsamuel@chromium.org>
Date: Mon Mar 19 15:44:53 2018

Display Scheduler: Don't request BeginFrames while waiting for a root surface

When the root surface changes, we cannot draw to the display until a surface
activates. This CL stops draws (and BeginFrames) until activation. If a root surface is never
submitted then the display will not listen for BeginFrames.

Bug:  821940 
Cq-Include-Trybots: master.tryserver.chromium.android:android_optional_gpu_tests_rel
Change-Id: I749c0ca8e85a3dbc7eccaa0799add81afb15beb9
Reviewed-on: https://chromium-review.googlesource.com/964595
Reviewed-by: Eric Seckler <eseckler@chromium.org>
Commit-Queue: Fady Samuel <fsamuel@chromium.org>
Cr-Original-Commit-Position: refs/heads/master@{#543548}(cherry picked from commit cfb855831a545f535565625209fe437017e05525)
Reviewed-on: https://chromium-review.googlesource.com/968901
Reviewed-by: Fady Samuel <fsamuel@chromium.org>
Cr-Commit-Position: refs/branch-heads/3359@{#304}
Cr-Branched-From: 66afc5e5d10127546cc4b98b9117aff588b5e66b-refs/heads/master@{#540276}
[modify] https://crrev.com/8adbb60e9662cc609447c691137aa88c38f71b8f/components/viz/service/display/display_scheduler.cc
[modify] https://crrev.com/8adbb60e9662cc609447c691137aa88c38f71b8f/components/viz/service/display/display_scheduler_unittest.cc

This should be in the next M66 beta. I'll follow up when we push out the new beta this week.

The current beta 66.0.3359.45 contains this fix. I'm marking this as FIXED.