Looking at the blame list:

https://codereview.chromium.org/2128533002
and 
 https://codereview.chromium.org/1735473002

are unlikely to be the reason.

the former is ios and the latter is ppapi.

I guess it is the skia roll then?
https://codereview.chromium.org/2125883002

I will speculatively revert the skia roll, since I cannot come up with any other reason.

also removing RVG.

landing the revert:
https://codereview.chromium.org/2122293002/

the revert really did not help. I will revert the revert.

The only clue I have is that if the screen is off then the test fails reliably when I locally run it.

This may be an infra issue.

The last swarming bot_config change was July 2, and the last swarming server change was June 30.

The failure mode does seem confined to swarming, though...

What if we rerun a swarming task from yesterday?

Sure, seems reasonable.

https://chromium-swarm.appspot.com/user/task/2fdaf416d8e06110 and https://chromium-swarm.appspot.com/user/task/2fdaf3fe422bcf10 are retries of the android_webview_test_apk and chrome_public_test_apk binaries built in https://uberchromegw.corp.google.com/i/chromium.linux/builders/Android%20Tests/builds/28628.

Both tasks finished successfully.

Speculatively reverting https://codereview.chromium.org/2122233003/ to see if that helps.

seems the revert worked. starting from https://uberchromegw.corp.google.com/i/chromium.linux/builders/Android%20Tests/builds/28645 build is green.

For debugging purposes, here are links to the failure logs and isolated outputs:

https://luci-logdog.appspot.com/v/?s=chromium%2Fbb%2Fchromium.linux%2FAndroid_Tests%2F28629%2F%2B%2Frecipes%2Fsteps%2Fandroid_webview_test_apk_on_Android%2F0%2Fstdout

https://isolateserver.appspot.com/browse?namespace=default-gzip&hash=4bca6d5ec625bf506904a5d82b14d56f7cf69bbf

https://chromium-swarm.appspot.com/user/task/2fd9bee861f0cc10