All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=697490

Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgxMfOuwkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpPalvAkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpLPgvQsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpPWR4ggM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpPWGvQkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpJ7hvAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpLPTpgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpKeorQkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpIDmtwkM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpODkvgoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpPzcvAsM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpPzIrQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpLy3tAoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgxOeErQoM,agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDgpJOBtgoM


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

webrtc-linux-large-tests
webrtc-win-large-tests

could be this: https://chromium-review.googlesource.com/c/443866/

The roll contains https://chromium.googlesource.com/webm/libvpx.git/+log/4d4231352c..8121f85473 so re #2: yes that's possible.

I don't see that we need to investigate improvements like this - looking at the improvements should only be to verify they're not "too good to be true", i.e. a metric like encode time that is normally at a certain amount of millisecond doesn't suddenly become 0 (which can be interpreted as a great improvement, but it's unlikely to be proper). If this is not clear in the perf docs - please update them!