> the third case the behavior is wrong, the `.cancel()` should win over the timeout. Otherwise using the context manager becomes too risky in real-world situations.

Please elaborate. The first code that calls `.cancel()` wins, doesn't matter what is the source. asyncio has no priorities.

> I also think your first graph has an issue if the user has a `try/except TimeoutError` between `timeout-a` and `timeout-b`, which is now more probable since we're dropping `move_on`. We can take the discussion to the forked repo; I can put together some tests if that would make it easier.

Ok, let's discuss on GitHub. I only would mention that no code could be executed between timeout-a and timeout-b, because both events are scheduled between the previous event loop iteration and the current one.
Sure, if we can start talking with code (and failed tests) -- it can raise the understanding level very much.