Layout test css3/filters/effect-reference-subregion-hw.html is failing flakily on the linux_chromium_rel_ng tryserver. This is causing valid tryjobs to be retried, slowing down the CQ. Some recent failures:

https://luci-milo.appspot.com/buildbot/tryserver.chromium.linux/linux_chromium_rel_ng/556508
https://luci-milo.appspot.com/buildbot/tryserver.chromium.linux/linux_chromium_rel_ng/556478
https://luci-milo.appspot.com/buildbot/tryserver.chromium.linux/linux_chromium_rel_ng/556420
https://luci-milo.appspot.com/buildbot/tryserver.chromium.linux/linux_chromium_rel_ng/556404
https://luci-milo.appspot.com/buildbot/tryserver.chromium.linux/linux_chromium_rel_ng/556376
https://luci-milo.appspot.com/buildbot/tryserver.chromium.linux/linux_chromium_rel_ng/554896

Originally seen as a flake on this CL:
https://chromium-review.googlesource.com/687374

The logs indicate that perhaps the problem's in the configuration of the X server or xvfb:

12:23:42.690 19752 worker/6 css3/filters/effect-reference-subregion-hw.html output stderr lines:
12:23:42.690 19752   
12:23:42.690 19752   DevTools listening on ws://127.0.0.1:57383/devtools/browser/589fc882-32b2-475a-b788-63b10e8f7b10
12:23:42.690 19752   Xlib:  extension "RANDR" missing on display ":100".
12:23:42.690 19752   [19998:19998:1002/122313.767018:2649617928:ERROR:gpu_info.cc(103)] No active GPU found, returning primary GPU.
12:23:42.695 18311 [11/11] css3/filters/effect-reference-subregion-hw.html failed unexpectedly (test timed out, text diff)
12:23:42.694 19752 worker/6 killing primary driver
12:23:42.694 19752 worker/6 killing secondary driver
12:23:42.694 19752 worker/6 css3/filters/effect-reference-subregion-hw.html failed:
12:23:42.695 19752 worker/6  test timed out
12:23:42.695 19752 worker/6  text diff

The test's marked "slow" but it's probably the case that it's intermittently timing out, not that the test is actually slow when it passes. A few old earlier bugs reference this test; referencing them.

To avoid disabling this test who can triage this? Thanks.