All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=917257

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=e9fab0103469f28d678806bc0e72de7a387c31886db2e95c41bc774cc8031e9b


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

Win 7 Nvidia GPU Perf

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

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/14e22b5e140000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/14e22b5e140000

[CrOS MultiDevice] Move appropriate CryptAuth files to SecureChannel. by khorimoto@chromium.org
https://chromium.googlesource.com/chromium/src/+/1ddc969449a5ccb60dd598b130dd33c701ae6c91
timeToFirstContentfulPaint: 1171 → 756.3 (-414.4)

[TaskScheduler]: Disable MergeBlockingNonBlockingPools from fieldtrial_testing_config. by etiennep@chromium.org
https://chromium.googlesource.com/chromium/src/+/ad3c94a2279bd7584003990858339a94f69a838b
timeToFirstContentfulPaint: 756.7 → 1053 (+296.2)

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

Benchmark documentation link:
  https://bit.ly/loading-benchmarks

Looks similar to  Issue 915769 .
It's related to disabling MergeBlockingNonBlockingPools feature from test config.
That feature was already disabled on stable, so it is not affected by the regressions.