All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=800004

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=4a5c0f6d1e8948c1230981acbb45bf16a09553be6ad75703d0668043b4ad4f2b


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

chromium-rel-win7-gpu-nvidia

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14e091db040000

📍 Couldn't reproduce a difference.
https://pinpoint-dot-chromeperf.appspot.com/job/14e091db040000

dtu: From the dashboard graphs, I really expected pinpoint to repro. Should I try to restart it? Is there a way to increase storyset-repeat?

The dashboard graph shows a jump from ~900 to ~2,800.
The pinpoint results show ~1,500 before and ~1,700 after. Definitely much smaller than the jump in the dashboard graph.

Everything looks right to me:
* The revision range looks right.
* The ending isolate hash is right.
* The benchmark across 4 different devices.

I don't think another run will produce a different result. I do want to note that this metric looks especially noisy in Pinpoint, with the histogram showing results that range from 0 to 3335. I'd suspect that some kind of small change in the testing environment caused the dashboard results to jump.

WontFixing because:
* bisect can't repro
* The metric is noisy in pinpoint
* It looks like the regression is just going back up after a temporary drop

+tdresser for noise in EQT metrics.