+Simon, Dave

So the bug only shows one regression: https://chromeperf.appspot.com/group_report?keys=agxzfmNocm9tZXBlcmZyFAsSB0Fub21hbHkYgIDg4YuZogoM

It's in API:duration_avg/http___meta.discourse.org

For some reason there are different pages being bisected on that bug, amazon and bbc. Only the first bisect fails to repro the regression on the linked graph.

Its numbers look like there was a significant increase in noise, which made the bisect stop? Simon, Dave, any ideas why it didn't keep bisecting? I'm going to try kicking another bisect on the page.

Revision             Result                  N
chromium@487550      9.479 +- 49.5735        21      good
chromium@487730      22.7316 +- 300.259      21      bad

FYI I manually added a few more regressions to the bug link https://chromeperf.appspot.com/group_report?bug_id=747799 from where I started the other bisects.

Ah, now I see more graphs in the bugs, I'm not sure why I didn't see them yesterday. Looking at the rest of the graphs, it looks like the jump goes up over 2-3 runs. I'll try kicking off some bisects over wider ranges to see if that's the issue.