All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=667675

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICgy9vo5wgM


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

webrtc-win-large-tests

erikchen@ : This regression seems possibly related to your CL https://codereview.webrtc.org/2496413002. Could you please take a look?

The CL in function is a functional change that only affects second monitors on Mac. Regressions are on linux/windows.

brandtr@: Your CL https://codereview.webrtc.org/2499273002 seems like a potential cause of this regression. Could you please take a look?

This is expected. Shortening of the the tests lead to fewer dropped frames, in absolute numbers. The decreased variance after the change might be due to what screenshare slides are actually shown: there is one "hard" slide which is probably only shown once after this change.