Thank you for reporting this.

I suspect this is caused by shortening delays in the tests (see  issue 650346 ).

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/9d2e15b7c11d53baa8226bad662b26663838e035

commit 9d2e15b7c11d53baa8226bad662b26663838e035
Author: nhiroki <nhiroki@chromium.org>
Date: Tue Dec 06 09:15:31 2016

Worker: Refine pending activity management on InProcessWorkerMessagingProxy

This CL...

(1) renames m_workerGlobalScopeMayHavePendingActivities to
    m_workerGlobalScopeHasPendingActivities. InProcessWorkerMessagingProxy
    cannot know the exact current state of pending activities on the worker
    context because it's in a separate thread, so the field name contained
    'may'. This would be confusing and pointless from caller's point of view.

(2) stops checking m_unconfirmedMessageCount in hasPendingActivity() in favor of
    m_workerGlobalScopeHasPendingActivities. When the count is greater than 0,
    m_workerGlobalScopeHasPendingActivities should be true.

(3) removes workerGlobalScopeMayHavePendingActivity() from
    InProcessWorkerMessagingProxyForTest because it's the same with
    hasPendingActivity() after (2).

Note that this CL does not change behavior.

BUG= 655089 

Review-Url: https://codereview.chromium.org/2555673002
Cr-Commit-Position: refs/heads/master@{#436549}

[modify] https://crrev.com/9d2e15b7c11d53baa8226bad662b26663838e035/third_party/WebKit/Source/core/workers/DedicatedWorkerTest.cpp
[modify] https://crrev.com/9d2e15b7c11d53baa8226bad662b26663838e035/third_party/WebKit/Source/core/workers/InProcessWorkerMessagingProxy.cpp
[modify] https://crrev.com/9d2e15b7c11d53baa8226bad662b26663838e035/third_party/WebKit/Source/core/workers/InProcessWorkerMessagingProxy.h

The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/d7babd6989862cc47c28874106f6dede10b059f5

commit d7babd6989862cc47c28874106f6dede10b059f5
Author: nhiroki <nhiroki@chromium.org>
Date: Tue Dec 06 09:25:48 2016

Worker: Fix test flakiness (timeout) on DedicatedWorkerTest

These tests wait for notifications from a worker context using nested loops. If
notifications are consecutively delivered, the balance between enterRunLoop()
and exitRunLoop() is lost and timeout happens.

This CL queues notifications from a worker context and makes sure to keep the
balance.

BUG= 655089 

Review-Url: https://codereview.chromium.org/2553973002
Cr-Commit-Position: refs/heads/master@{#436552}

[modify] https://crrev.com/d7babd6989862cc47c28874106f6dede10b059f5/third_party/WebKit/Source/core/workers/DedicatedWorkerTest.cpp