UserAgent: Mozilla/5.0 (Windows NT 10.0; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/67.0.3396.103 YaBrowser/18.7.0.2695 Yowser/2.5 Safari/537.36

Steps to reproduce the problem:
The problem appears for example in media.tough_video_cases

What is the expected behavior?

What went wrong?
There are some Telemetry issues containing stack traces like this:

[ RUN      ] video.html?src=garden2_10s.ogv
Traceback (most recent call last):
  File "C:\b\depot_tools\python276_bin\lib\SocketServer.py", line 593, in process_request_thread
    self.finish_request(request, client_address)
  File "C:\b\depot_tools\python276_bin\lib\SocketServer.py", line 334, in finish_request
    self.RequestHandlerClass(request, client_address, self)
  File "C:\b\depot_tools\python276_bin\lib\SocketServer.py", line 651, in __init__
    self.finish()
  File "C:\b\depot_tools\python276_bin\lib\SocketServer.py", line 710, in finish
    self.wfile.close()
  File "C:\b\depot_tools\python276_bin\lib\socket.py", line 279, in close
    self.flush()
  File "C:\b\depot_tools\python276_bin\lib\socket.py", line 303, in flush
    self._sock.sendall(view[write_offset:write_offset+buffer_size])
error: [Errno 10053] An established connection was aborted by the software in your host machine

They are unrelated to these issues and they are distracting. 
Moreover, I noticed _MemoryCacheHTTPServerImpl stops handling requests on rare occasions (I can't reproduce this on chrome, it happens in our forked tests) and for some reason handling errors fixes the problem.

Did this work before? N/A 

Chrome version: dev  Channel: n/a
OS Version: 10.0
Flash Version: Shockwave Flash 31.0 r0