The following revision refers to this bug:
  https://chromium.googlesource.com/chromium/src.git/+/f9afa68c5c296d11ad1627780e44f02699df7ab5

commit f9afa68c5c296d11ad1627780e44f02699df7ab5
Author: Min Qin <qinmin@chromium.org>
Date: Fri Jan 18 23:10:50 2019

Fix an issue that browser is launched when resuming a download with service manger

onStartTaskWithNative() was called later than supportsServiceManagerOnly().
As a result, mCurrentTaskType is not initialized when supportsServiceManagerOnly() is called.
Move the initialization to onStartTaskBeforeNativeLoaded().

BUG=923522

Change-Id: I27b590b634f72d08e9ce672717ef40f378dec773
Reviewed-on: https://chromium-review.googlesource.com/c/1422441
Reviewed-by: Shakti Sahu <shaktisahu@chromium.org>
Commit-Queue: Min Qin <qinmin@chromium.org>
Cr-Commit-Position: refs/heads/master@{#624343}
[modify] https://crrev.com/f9afa68c5c296d11ad1627780e44f02699df7ab5/chrome/android/java/src/org/chromium/chrome/browser/download/service/DownloadBackgroundTask.java

Curious - how was this detected? Don't we have a test that should fail and thus blocked the original commit?

I was testing something else in reduced mode manually, and found it had been broken for a couple of days.

I think the reason our tests did not catch it is that full browser was being started ~1.5s after reduced mode started, so the test had already passed by then.

Ugh. I want to say a perf test might be more likely to catch this (say if memory suddenly shot up) but has the same challenge of potentially ending too early