In terms of performance, behaves identically to core.async so no regression here:

but promesa impl does not have the limitation of 1024 watchers, so I can run the same example with more processes without any problems

https://gist.github.com/niwinz/2c08cc0366c943207e9ae0e70ac52cf3