All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=793238

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=3b71419d0142876a7d12e8c999f4f1371f067165e422a73b5e7db1b15637ecf5


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

chromium-webrtc-trunk-tot-rel-win7

A few observations:
- The regression hits only windows bots, but all of them (win7, win8, win10).
- The alert fired only for FYI bots, but the same thing happens for non-FYI bots. This would indicate that this is not due to a WebRTC ToT CL.
- The intersection of all blame lists is: https://chromium.googlesource.com/chromium/src/+log/e39107d2d90d21e659c377e7b05599fa2957f79b..bacb06c84b2ae042242196cd4e510bff20a02999

Two culprits in this range:

1. WebRTC roll at https://chromium.googlesource.com/chromium/src/+/972e7900aa71d451fe1da7c64ff486e22b9256b6. I don't think this is the culprit, since the FYI bots should have triggered when these WebRTC changes were landed on WebRTC ToT; they didn't.

2. Make WebRTC's rtc::Event implementation rely on WaitableEvent. by Tommi, https://chromium.googlesource.com/chromium/src/+/cda8addd5b0934e2ce45762f7b43ede30ed7c12b.

You can see more regressing bots at https://chromeperf.appspot.com/report?sid=23e0dac93b9ccf321d61d89e5ed0b5dd1fadcc5fe5aac0c885e5c83122b260b7&start_rev=521290&end_rev=522738.

Tommi, I suspect your CL caused this regression. Is this expected?

I'm expecting some changes, yes but wasn't sure where they'd pop up. Do you
know if the affected code does frequent locking (e.g multiple small scoped
locks vs holding the lock longer) and if contention is common?

Looking closer, it appears that this is actually an improvement. The implementation we're now using in Chromium, is more accurate than the default implementation in WebRTC and the wait periods are now less busy.

Comparing the windows bots to the linux and mac bots, we're now getting consistent results across those platforms.  See e.g.:

https://chromeperf.appspot.com/report?sid=c3895c395e4a6b74a26c61ed328beaf5eb401d0e747b418c763340f10cd6ec71

Issue 793320 has been merged into this issue.