All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=593603

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6nqsQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6n-qAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_7CSuQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA__7asgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggKClqQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_8vAqgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA__6BqwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggKCzuAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_63VpgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6mluQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggKDPpQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_8uUuwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggODYtAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6LVtgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6KLrwoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6KfvgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_8u3oQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgICggKCzuAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6uosAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA__qPpQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_4HdqAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_6uNtAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_5KMvQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDA_5SFsQoM


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

webrtc-android-tests-nexus5
webrtc-android-tests-nexus9
webrtc-linux-large-tests
webrtc-mac-large-tests
webrtc-win-large-tests

A likely CL that caused this could be https://codereview.webrtc.org/1773743002.

Reassigning to tkchin@

The above CL link was the wrong one. Reassigning back to peah@

Reassigning to pbos@
Could you please have a look to see whether this could be related to https://codereview.webrtc.org/1780543003

Not related, the code isn't hit in those tests. marpan@ the suspect CL rolls a new libvpx from Chromium, can you take a look?

Suspect CL: https://chromium.googlesource.com/external/webrtc/+/7e749949776f9da58e8a1a8815eaa638c6510198

It looks like a regression/bug in libvpx regarding scene/slide change detection, which got into the roll mentioned in #7. A fix has been merged into libvpx and should be rolled in next week. Will monitor the graphs after the roll.

The issue only affects vp9 under screensharing.

This still hasn't recovered, though none of the individual pages appear to have regressed at all.

Filed infrastructure issue here: https://github.com/catapult-project/catapult/issues/2151

Resolution to the infra bug:

The reason why bisect is disabled for those tests is because we don't have bisect bots set up to run the WebRTC perf tests.

screenshare_slides_vp9_2sl isn't an aggregate of the other series listed in the legend in those graphs.

marpan, re comment 8: graphs do seem much improved. Should this be marked fixed?

Yes the improvement corresponds to the libvpx fix rolled into chromium this week. Should be marked as fixed.