Oleg Pyzhcov

2.13.5 has added linting for unused context bounds ([this pull](https://github.com/scala/scala/pull/9346)).

Wow. This is crazy. And thanks for the report, btw. Seems to be caused by an odd interaction with `ApplicativeLocal` inheriting `ApplicativeAsk` (this is the only typeclass using inheritance instead...

@kubukoz care to provide more details?

FWIW fs2 behavior when forcing rechunking (closest thing to our buffering) isn't safe either: https://scastie.scala-lang.org/5h9bshLyS9uiyTOop9a8wQ I also think that monix having a _variation of Reactive Streams protocol_ can allow us...

I'm pretty sure the "proper" way to solve your original issue is to do ``` t .executeOn(s /*, forceAsync = true*/) //

> What are your thoughts about `runAsync` ? I feel like these two should be consistent I feel like `runAsync` is API for masochists that you avoid like plague if...

So the plan is to keep the default value but not as implicit, or also to remove the default value?

`tailRecM` is never required to be _memory_ safe for infinite loops by `Monad`, and you shouldn't use it for something that's not an effect type _only_. `cats.effect.Resource` extends `StackSafeMonad` yet...

> isn't memory safety the whole point of tailRecM? Stack only. Not heap safety. [The paper](https://functorial.com/stack-safety-for-free/index.pdf) says the law to satisfy is: >A valid implementation of `MonadRec` must guarantee that...

> IMO that distinction isn't useful, being an implementation detail. A language like Haskell doesn't even have a stack, being all heap. It's literally the problem being solved, and the...