Hi, Not sure if this a bug, or an intended feature. 
I got surprise by the following behavior. 


    from asyncio import run, sleep, get_event_loop

    print(get_event_loop()) # return the current eventloop
    run(sleep(0))
    print(get_event_loop()) # raise a RuntimeError


I would expect `get_event_loop` to get back to it's initial default value. 

This comes from the fact that `run()` call `set_event_loop()` to `None`. with
both sets `_set_called` to `True`, but `get_event_loop` seem to assume if
_set_called is True, then loop cannot be none. 

I'm tempted to think that if `set_loop()` is called with `None`, it should reset the `_set_called` to False. Or Am I supposed to call `set_event_loop(new_event_loop())` myself ? 

I'm likely missing  something so any insight would be appreciated; if you
believe this is an actual issue I'm happy to send a PR.

I think now it is too late for changing the behavior.
In the ideal world, get_event_loop() should never exist, maybe we can deprecate/remove it eventually.
run() should be used for executing async code; get_running_loop() for getting a reference to the currently executed loop.
It may help to avoid many silly problems.
Maybe some scenarios are not covered but at least high-level code can be organized this way.

P.S.
set_event_loop(None) is not for resetting to default but for disabling of the default loop creation for the current thread. Non-main threads work like set_event_loop(None) was called implicitly.

Thanks Andrew; 

I'd still would love to have a way to get the current eventloop outside of async code that is consistent. 

I guess throwing an error with `get_event_loop()` is ok, it's just that `RuntimeError` feel like a weird thing to catch.

Alternatively is the current event loop is None I feel like returning None would also make sens imho. 

Feel free to close if you think this will not get fixed/updated.

Asyncio has a few own exceptions, the library tries to utilize existing ones, e.g. ValueError, TypeError, and OSError family. RuntimeError falls into this category.

get_event_loop() never returns None, it was from very beginning. If we change it existing third-party code will be broken. So, no way, sorry.

I really don't think we can change anything here.

Would you describe cases when you need an implicit loop instance but there is no running loop currently?

The usage of get_event_loop() outside of a loop is deprecated since Python 3.10