The test output from test_sys_settrace makes it look like a couple of the async tracing tests aren't cleaning up after themselves properly:

```
[ncoghlan@localhost cpython]$ ./python -m test test_sys_settrace
Run tests sequentially
0:00:00 load avg: 1.27 [1/1] test_sys_settrace
unhandled exception during asyncio.run() shutdown
task: <Task finished name='Task-8' coro=<<async_generator_athrow without __name__>()> exception=RuntimeError("can't send non-None value to a just-started coroutine")>
RuntimeError: can't send non-None value to a just-started coroutine
unhandled exception during asyncio.run() shutdown
task: <Task finished name='Task-11' coro=<<async_generator_athrow without __name__>()> exception=RuntimeError("can't send non-None value to a just-started coroutine")>
RuntimeError: can't send non-None value to a just-started coroutine

== Tests result: SUCCESS ==

1 test OK.

Total duration: 102 ms
Tests result: SUCCESS

```

If that output is actually expected as part of the test, it would be helpful if the test printed a message beforehand saying to expect it. Otherwise, it would be desirable for the test to clean up after itself and keep the messages from being displayed in the first place.

It looks like setstrace function trigger this codeline: https://github.com/python/cpython/blob/master/Modules/_asynciomodule.c#L2618

Looks we can use set_exception_handler() to catch this error, but i am not sure it is worth catching this error through adding  a extended exception hander.

so, Nick, what's your thought?

REF: https://github.com/python/cpython/blob/master/Lib/asyncio/base_events.py#L1627

Once consensus is reached, I would be happy to do the work.

I'm not seeing this output when I run the test. Are you still seeing it?

I'm not, so I assume this got cleaned up somewhere along the way.