I'm not sure what you mean by "multi-threaded async/wait"... Isn't the article considering async/await as an alternative to threads (i.e. coroutines vs threads)?
I'm a C++ programmer, and still using C++17 at work, so no coroutines, but don't futures provide a similar API? Useful for writing async code in serialized fashion that may be easier (vs threads) to think about and debug.
Of course there are still all the potential pitfalls that you enumerate, so it's no magic bullet for sure, but still a useful style of programming on occasion.
They mean async/await running over multiple OS threads compared to over one OS thread.
You can also have threads running on one OS thread (Python) or running on multiple OS threads (everything else).
Every language’s concurrency model is determined by both a concurrency interface (callbacks, promises, async await, threads, etc), and an implementation (single-threaded, multiple OS threads, multiple OS processes).
async/await tasks can be run in parallel on multiple threads, usually no more threads than there are hardware threads. This allows using the full capabilities of the machine, not just one core's worth. In a server environment with languages that support async/await but don't have the ability to execute on multiple cores like Node.js and Python, this is usually done by spawning many duplicate processes and distributing incoming connections round-robin between them.
I'm a C++ programmer, and still using C++17 at work, so no coroutines, but don't futures provide a similar API? Useful for writing async code in serialized fashion that may be easier (vs threads) to think about and debug.
Of course there are still all the potential pitfalls that you enumerate, so it's no magic bullet for sure, but still a useful style of programming on occasion.