All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=872168

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=ce468a53ba4b4bb49b4eb47d31c7b7fce633df01028635771fdbd7da0eff8023


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

Android Nexus6 WebView Perf

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/11cf6da4640000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/11cf6da4640000

Roll AFDO from 70.0.3512.0_rc-r1 to 70.0.3513.0_rc-r1 by afdo-chromium-autoroll@skia-buildbots.google.com.iam.gserviceaccount.com
https://chromium.googlesource.com/chromium/src/+/fec0224315eaaeb75b118b9b3705bad2423c48c6
1248 → 815.2 (-432.9)

Assigning to sheriff gbiv@chromium.org because "Roll AFDO from 70.0.3512.0_rc-r1 to 70.0.3513.0_rc-r1" is a roll.

Understanding performance regressions:
  http://g.co/ChromePerformanceRegressions

Looks like this is basically undoing the perf gains from the 3512 profile roll.

3512 pretty seriously regressed bits of rasterize_and_record for both linux and nexus 5(x?) bots. So I'm betting that 3512 was just a wacky profile. Since there's no other obvious afdo-related flakes in (admittedly short) history here, all I can really say is "¯\_(ツ)_/¯".