All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=701739

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDglMOlswsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1K7O7woM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg1KH_nQkM


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

webrtc-mac-large-tests
webrtc-win-large-tests

Adding a new full stack test (https://codereview.webrtc.org/2745523002) seems to affect other tests, even on Mac and Windows.

Have you seen this before, ilnik?

I've seen that behavior before. Adding one video test broke some audio test. The funny thing is: repeating audio test several times resolved issue: only first run were worse.

All changes are very minnor (e.g. 2ms change in receiver_time at screenshare_slides_scroll test). Also, they are not accompanied by correlated metrics (e.g. cpu_usage did not change). We can assume it's only a noise or hardware/architecture problems. Change of size of the executable could be blamed for all these changes alone.