All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=773258

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=4fbe0e19aec69fec8b498a56fc64852a950773fa5a670194462b46f3316051c6


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

webrtc-android-tests-nexus5-kitkat

On the graphs, this looks like a change to variance more than to the average. For android, and "largeroom_5thumb" tests (and one alert also for conference_motion_hd_2tl_moderate_limits.

Rasmus, do you know these tests, and are the apparent change in variance normal? The different behavior appeared around commit #20127, and disappeared around commit #20172, but I find no likely culprit cls. Magnus, do you now any recent android changes which amy be related?

I'm tempted to close as WONTFIX, since changes are small and "back to normal", but I'd like to hear your advise first.

Agree this can be WontFix'd. The Nexus5-kitkat bot is noisy, and if there is no obvious culprit CL there's no point spending more time on this.