All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=903160

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=94907e47b890ff8472d3af9959fd59cf1fcb149bbc3ea64d72569c410e5942d2


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

Android Nexus5 Perf
android-nexus5x-perf

system_health.common_mobile - Benchmark documentation link:
  https://bit.ly/system-health-benchmarks

v8.browsing_mobile - Benchmark documentation link:
  None

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/15b77b2be40000

📍 Found a significant difference after 1 commit.
https://pinpoint-dot-chromeperf.appspot.com/job/15b77b2be40000

[Presentation API] Remove from insecure contexts. by mfoltz@chromium.org
https://chromium.googlesource.com/chromium/src/+/592f6faa792259bb3515fea429699c3e126a143e
console:error:all: 9.7 → 14.8 (+5.1)

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

Benchmark documentation link:
  https://bit.ly/system-health-benchmarks

I am digging and trying to find out how this test works.   It seems clear from the initial investigation that whatever page is being loaded as part of the story, is not currently being served by techcrunch.tumblr.com.

I spent quite some time digging around.  The only clue that is possibly related to my change is that there are console messages logged from a YouTube domain related to an attempt to load the Cast SDK.  It's possible that the archived pages (that appear to be used by this test) happened to be grabbed while YouTube was running experiments to support Cast on mobile Web in 2016 (?).  It's trying to load Chrome extension resources on Clank, which is just wrong.

In fact, YT has deprecated and removed support for the Cast SDK, so users would not see this today.  If I can find the place with the archive, I can confirm this hypothesis.

I wasn't able to reproduce any of this regression with http://techcrunch.tumblr.com.  In fact, the blog hasn't been updated for 2 years, so I am not sure this is a good test case for infinite scrolling - you can't scroll very far before reaching the end of the content.

After some searching, I found a tumblr with YouTube embeds (http://nprmusic.tumblr.com) that did not log any of the console errors that were responsible for this regression.  It's got a lot more content, so I'm happy to update the test case if you think that's a good path forward.

Let me know if you have thoughts on how to proceed.

I'm closing based on lack of feedback to my last comment.  If anyone wants to pursue updating this test, feel free to pick this up.