I saw a few test timeouts on the waterfall earlier today - the Maps pixel test timing out and a couple of other WebGL conformance tests as well.

There wasn't an obvious culprit but something has definitely regressed. My best guesses would be either V8 or the loader. Can anyone try to come up with even a conservative blamelist based on the timeouts seen on the waterfall and the relative duration between them, going back in time to a point well before the first failure?

Maps timeouts seen earlier:

https://ci.chromium.org/buildbot/chromium.gpu.fyi/Win7%20x64%20Debug%20%28NVIDIA%29/8811
https://ci.chromium.org/buildbot/chromium.gpu.fyi/Win7%20x64%20Debug%20%28NVIDIA%29/8806

I've looked at several builds on Mac Retina Release (NVIDIA) and the time varies significantly for these 2 tests between builds. One build it will be 290 seconds, the other 70 seconds. Maybe it's affected by garbage collection?

Maps used to time out before, too:
https://ci.chromium.org/buildbot/chromium.gpu.fyi/Win7%20x64%20Debug%20%28NVIDIA%29/8707

When it succeeds, it's not even close to 320 sec timeout.
In https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20x64%20Debug%20%28NVIDIA%29/builds/8813 it's 67 sec.
In https://build.chromium.org/p/chromium.gpu.fyi/builders/Win7%20x64%20Debug%20%28NVIDIA%29/builds/8810 it's 89 sec.

Doesn't seem like this is a case of a flaky bot either.
I think someone needs to try to repro and then see what causes the timeout.

Maybe connected to  https://crbug.com/v8/6927  ? CC mem sheriff.

Is there a way to rerun these tests on swarming? E.g. execute only the timed-out tests withe some repeat parameter?

Hmm, maybe. This is the full browser, though, not a gtest harness. If there's a general bug in process launching or teardown it could be related. Yes, it's certainly possible to manually trigger jobs on Swarming even against your own local builds; the maps_pixel_test might be a good candidate because it's just a single test. vmiura@ has done this and might be able to give you a pointer tomorrow.

Note also that we're seeing tests time out on both Windows and Mac, and that the Mac GPU bots (the Mac Minis with Intel GPUs and MacBook Pro Retinas with AMD GPUs) are all running 10.12.

Given the investigation in  https://crbug.com/v8/6927  it seems likely this was the same bug. Please reopen or file a new bug if hangs persist.