All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=867041

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=926c9cc7d5c03a93328492a1ef41372498f353cea1e22ec496a01e2147e151eb


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

mac-10_13_laptop_high_end-perf

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/17f1248ba40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/17f1248ba40000

Get page-scale-factor from RenderFrameMetadata in hit-testing tests. by riajiang@chromium.org
https://chromium.googlesource.com/chromium/src/+/a26a8b8af90fde922e9eac9cec747d6160b5c9ef
264.7 → 307 (+42.29)

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

https://chromium.googlesource.com/chromium/src/+/a26a8b8af90fde922e9eac9cec747d6160b5c9ef is only affecting tests (i.e. observe RenderFrameMetadata in tests) so I don't think this is the culprit CL. These bots are also mac specific.

+jonross@ to confirm that the observation of RenderFrameMetadata in tests shouldn't affect these metrics.

The change in #3 only affects testing, and would not show up in page loading on Mac.

Looking at the pinpoint in #3 the blamed change's error bars overlap with those of other revisions.

A few next steps:
  - redo the pinpoint
  - Expand the bisect range on the dashboard in #2
  - Look at the changes in the other revisions with giant error bars in the pinpoint

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/1483fc9ba40000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/12e466dda40000

Thanks Jon! I restarted the pinpoint job.

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/12e466dda40000

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/11ab2163a40000

The graph goes up over multiple runs. Trying the bisect on a wider range.

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/11ab2163a40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/1483fc9ba40000

Get page-scale-factor from RenderFrameMetadata in hit-testing tests. by riajiang@chromium.org
https://chromium.googlesource.com/chromium/src/+/a26a8b8af90fde922e9eac9cec747d6160b5c9ef
261.5 → 327 (+65.52)

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

Pinpoint runs with pre-built binaries, so I think when it's pointing to riajiang's change multiple times it's just detecting that the binary is different somehow, not that there's something specific to that change that causes pages to load more slowly.

It looks like this is really not reproducing, but +tdresser in case we should investigate noisiness of metrics.