We have a 5-10% failure rate for logdog bootstrap in the WebRTC waterfall for one specific bot - vm828-m1.labs. We don't see it for other bots for some reason. Is it because of the win filesystem or some characteristic of the lab network?

https://ci.chromium.org/buildbot/chromium.webrtc/Win%20Builder/?limit=200

Example failure:

https://ci.chromium.org/buildbot/chromium.webrtc/Win%20Builder/81935

Traceback (most recent call last):
  File "../../../scripts/slave/remote_run.py", line 786, in main
    return _main_impl(argv, stream)
  File "../../../scripts/slave/remote_run.py", line 776, in _main_impl
    cleanup_dir, cache_dir)
  File "../../../scripts/slave/remote_run.py", line 677, in _exec_recipe
    recipe_return_code = bs.get_result()
  File "E:\b\build\scripts\slave\logdog_bootstrap.py", line 661, in get_result
    self._bootstrap_result_path, e))
BootstrapError: Failed to open bootstrap result file [E:\b\rr\tmpm81nze\bootstrap_result.json]: [Errno 2] No such file or directory: 'E:\\b\\rr\\tmpm81nze\\bootstrap_result.json'