All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=845875

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=62781937b5325c0c9d948cbdc1e06f15bc758b5fd511123facc2164a37c7e374


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

webrtc-win-large-tests

Main suspect for this change in memory usage is the chromium roll https://webrtc-review.googlesource.com/75020, including libvpx changes https://chromium.googlesource.com/webm/libvpx.git/+log/e4408a07be..28801f91c4

Marco, can you have a look?

Since it seems to be only one windows bot, I guess it might also be some bot config change. Patrik, are you aware of any bot changes?

Nope, you can click through to buildbot status page to see that the same machine executed both tests (vm347-m3). There has been no reconfiguration either AFAIK.

Not sure what could be causing the regression on this screenshare test, but there have are been a number of (positive) screenshare changes in vp9 recently. Since this regression is small and only on windows, will mark as won't fix.