Assigning to charliea@chromium.org as owner of mmutex.

I'll dig into this.  Thanks for sending it my way.

iannucci@, do you know if I can look at these on any flakiness dashboard? It'd be good to know if there are other flakes.

So the goal of the test that's flaking is to verify that the drain file is cleaned up as soon as the lock is acquired (and before the command is executed), rather than waiting until after the command has finished executing.

Digging around, I also see some flakes of this on Mac. I think that I've found a way to rewrite this test so that we're not just waiting some arbitrary amount of time for the command to start executing, and I'll send out a CL with this rewrite momentarily.

The following revision refers to this bug:
  https://chromium.googlesource.com/infra/luci/luci-go.git/+/01fe0de73b8b216d62f595a7c552bd8dcc622cda

commit 01fe0de73b8b216d62f595a7c552bd8dcc622cda
Author: Charlie Andrews <charliea@chromium.org>
Date: Thu Dec 14 19:27:03 2017

Fix flakiness in a runExclusive test by using channels instead of waits

This test has been flaky on the Mac and Win trybots (~3 flakes in ~15
days is my best guess).

Bug:  794666 
Change-Id: I86995ccf5f2e07c297b41fddbebe7e8ef297a3ec
Reviewed-on: https://chromium-review.googlesource.com/826411
Reviewed-by: Nodir Turakulov <nodir@chromium.org>
Commit-Queue: Charlie Andrews <charliea@chromium.org>

[modify] https://crrev.com/01fe0de73b8b216d62f595a7c552bd8dcc622cda/mmutex/lib/exclusive_test.go

Let me know if you see any further issues with this.