All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=883805

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=ec8d8b4d3e10b8349a7678801bfa392460472536e83bea7ccbb882da947f7144


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

linux-perf

power.desktop - Benchmark documentation link:
  https://bit.ly/power-benchmarks

Assigning to ekaramad@chromium.org because this is the only CL in range:
Reland "Implement MimeHandlerViewFrameController"

The initial attempt was causeing build failures in "mac-rel" bot.

The cause seems to be the use of static initializers. This CL
moves them to a method.

This is a reland of 92df23c1ce15615ba52698264485e8785103492c

Original change's description:
> Implement MimeHandlerViewFrameController
>
> This CL implements the renderer side of MimeHandlerView based on
> cross-process frames (behind a flag).
>
> The new subclass of MimeHandlerViewContainerBase will use a content
> frame inside a plugin element for loading the MimeHandlerViewGuest's
> extension.
>
> The current implementation is not complete and the missing features
> such as postMessage support will be added in the follow-up CLs.
>
> This CL also introduces a new class of parametric browser tests
> (MimeHandlerViewCrossProcessTest) which will be the destination of
> the current MimeHandlerViewTests as more features are integrated into
> the frame-based version. The new test class runs the test on both
> BrowserPlugin-based and cross-process-frame-based versions.
>
> Document: https://docs.google.com/document/d/10g7Y9cprYKkch9JZ0TBUWaEnHBJT1nzhskQIt1nHbWM/edit#heading=h.qw81a2bk3v6w
>
> Bug: 659750
> TBR=thestig@chromium.org
>
> Change-Id: Ia69aeed47f4fa1c7f5b81cdde71a8d8b5ff59165
> Reviewed-on: https://chromium-review.googlesource.com/1187231
> Commit-Queue: Ehsan Karamad <ekaramad@chromium.org>
> Reviewed-by: Sam McNally <sammc@chromium.org>
> Reviewed-by: Ehsan Karamad <ekaramad@chromium.org>
> Reviewed-by: Alex Moshchuk <alexmos@chromium.org>
> Reviewed-by: Istiaque Ahmed <lazyboy@chromium.org>
> Reviewed-by: James MacLean <wjmaclean@chromium.org>
> Cr-Commit-Position: refs/heads/master@{#589284}

TBR=alexmos@chromium.org,lazyboy@chromium.org,sammc@chromium.org,wjmaclean@chromium.org

Bug: 659750,  881696 
Change-Id: I94c779664981738222c8d79ca0b10ab06174a4a7
Reviewed-on: https://chromium-review.googlesource.com/1213369
Reviewed-by: Lei Zhang <thestig@chromium.org>
Reviewed-by: Ehsan Karamad <ekaramad@chromium.org>
Commit-Queue: Ehsan Karamad <ekaramad@chromium.org>
Cr-Commit-Position: refs/heads/master@{#589807}

The CL above implements a certain feature "behind flag". I do not see how this could have caused any perf regressions at all.

Could the range possibly be incorrect?

What kind of flag? I think perf bots run under some flags, not familiar with which. Taking a look at the graphs, it looks like there was something going on with the metric itself because the ref graphs moved in the same way. Marking WontFix.

The flag is --enable-features=MimeHandlerViewInCrossProcessFrame and I doubt bots would use that.