All graphs for this bug:
  https://chromeperf.appspot.com/group_report?bug_id=886661

(For debugging:) Original alerts at time of bug-filing:
  https://chromeperf.appspot.com/group_report?sid=aa313be4291e6a45d40052f475312469249089af6feead66e2bb322ad6e94c8e


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

Android Nexus5 Perf

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

📍 Pinpoint job started.
https://pinpoint-dot-chromeperf.appspot.com/job/16def23f640000

📍 Found significant differences after each of 2 commits.
https://pinpoint-dot-chromeperf.appspot.com/job/16def23f640000

[config] Enable scheduled script streaming field trial by leszeks@chromium.org
https://chromium.googlesource.com/chromium/src/+/7eb2bb6769b38736fc75a12c15563c4a69f1a61f
2.569e+07 → 2.603e+07 (+3.443e+05)

[BrowserSwitcher] Add Linux support by nicolaso@chromium.org
https://chromium.googlesource.com/chromium/src/+/9a4a94d66c8b42c979b083a4062ea1b696b146e8
2.603e+07 → 2.64e+07 (+3.666e+05)

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

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

It seems unlikely that my CL caused this regression, since the code only compiles on Windows and desktop Linux. This looks like a regression on Android.

Assigning to leszeks@ for him to take a look.

Correct, I've observed this flag flip increasing mobile memory use.