pfeldman@, I wonder if you could hind an owner for this bug?  This seems to have the same symptoms as  issue 595200  from 2 weeks ago:

https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/197823/steps/start_crash_service/logs/stdio:

Traceback (most recent call last):
  File "E:\b\build\scripts\slave\chromium\run_crash_handler.py", line 71, in <module>
    sys.exit(main())
  File "E:\b\build\scripts\slave\chromium\run_crash_handler.py", line 50, in main
    raise chromium_utils.PathNotFound('Unable to find %s' % exe_path)
common.chromium_utils.PathNotFound: Unable to find E:\b\build\slave\win\build\src\out\Release\crash_service.exe

Oh, and my apologies for throwing a more-or-less random bug at you - I found your name if history of changes for scripts/slave/chromium/run_crash_handler.py - this is why I hope you might know who can investigate this bug further.

Detected 3 new flakes for test/step "start_crash_service". To see the actual flakes, please visit https://chromium-try-flakes.appspot.com/all_flake_occurrences?key=ahVzfmNocm9taXVtLXRyeS1mbGFrZXNyHgsSBUZsYWtlIhNzdGFydF9jcmFzaF9zZXJ2aWNlDA. This message was posted automatically by the chromium-try-flakes app. Since flakiness is ongoing, the issue was moved back into Sheriff Bug Queue (unless already there).

I'm guessing this is something for the Trooper to look at:

common.chromium_utils.PathNotFound: Unable to find E:\b\build\slave\win\build\src\out\Release\crash_service.exe

The issue had an owner, and thus didn't show properly in the trooper queue.

As today's trooper, I'll take a look, but my hunch is I won't be able to do much for a missing executable.

It appears the step ran successfully on win_chromium_rel_ng since Apr 4

E.g. https://build.chromium.org/p/tryserver.chromium.win/builders/win_chromium_rel_ng/builds/202696

It also hasn't failed since Apr 8 globally. The step seems to fail in bursts about once a week. Not sure what to do about it, since the logs are now gone. 

Sample Dremel query to confirm:

SELECT
  ANY(master) sample_master, ANY(builder) as sample_builder, ANY(build_number) as sample_build, COUNT(build_number) as num_builds, host_name,
  FORMAT_UTC_USEC(TIME_USEC_TO_DAY(build_sched_msec * 1000)) as timestamp_utc
FROM chrome_infra.completed_steps
WHERE
  step_name == 'start_crash_service'
  AND result IN ('INFRA_FAILURE', 'FAILURE')
  -- AND builder = 'win_chromium_rel_ng'
GROUP BY host_name, timestamp_utc, 
ORDER BY timestamp_utc DESC, num_builds DESC
LIMIT 1000

Now that we have logdoc with a permanent log retention, I'd like to wait for a few more days to see another instance of a failure.

This step should be gone now everywhere, I think (we're getting rid of crash_service as it's not actually needed).