Alex pointed out

"""
Clicking through to the missing shard log, I see the following error:
[1031/162139:ERROR:test_launcher.cc(841)] Failed to read the filter file.
So I'm guessing something broke with how our .filter files are passed to the bot with swarming.
"""

that seems like it's probably the cause.

Thanks for looking into this Alex and helping identify the culprit.

git blame says that both lines referring to testing/buildbot/filters/isolate-extensions.*.filter come from r428848, so let me just try to revert this CL to unblock linux_site_isolation bot on the CQ (the CL being reverted was only trying to switch the Windows bot from --site-per-process to --isolate-extensions testing - reverting it shouldn't be a problem AFAICT).  FWIW, the revert is churning through the tryjobs at https://codereview.chromium.org/2466933003.

FWIW, the linux_site_isolation tryjob for the revert is @ http://build.chromium.org/p/tryserver.chromium.linux/builders/linux_site_isolation/builds/4437

I'm guessing the purpleness itself is caused by the fact that we add site-per-process filters as a dependency so that they're picked up for swarming, but we don't do that for isolate-extensions filters [1].  So the bot wasn't able to find the isolate-extension filter file once it was unexpectedly switched to --isolate-extensions, and the revert should hopefully fix it.

That means before relanding r428848, we should similarly add isolate-extensions.*.filter to deps, since swarming is now enabled on all our bots.

[1] E.g., https://cs.chromium.org/chromium/src/chrome/test/BUILD.gn?q=site-per-process.*.filter&sq=package:chromium&l=2018&dr=C

The purpleness is fixed starting in build https://build.chromium.org/p/chromium.fyi/builders/Site%20Isolation%20Linux/builds/12293

Unfortunately the purple has changed to red because of a new layout test.  I've opened  issue 661725  to track this.  I believe the new failure shouldn't impact CQ because the test will fail with and without a patch.