Why is it problematic to have tasks automatically get cancelled at suspension points? As I mentioned in the structure concurrency thread this is what Scala’s Zio library does and they consider it a feature. It isn’t clear to me why you would want to allow a task to continue across a suspension point once it has been cancelled already.
Sometimes it is necessary to prevent cancellation but I think that is the wrong default. Relying exclusively on cooperative cancellation seems to me like a good way to waste resources.
I’m sure you’ve thought this through in great detail but the tradeoffs and rationale are not presented in the proposal so it’s hard for me to know whether I’m missing important details or whether we just have different opinions about the right design decision. It would be helpful if you can spell this out in more detail.