Isn't Applicative strong enough for MonadComprehensions? Why not?

Question

It strikes me that every monad comprehension ends in a return. Does that not effectively make them isomorphic to applicative programming? Why do we have monad comprehensions with a Monad constraint instead of applicative comprehensions with an Applicative constraint?

This do-like notation for Applicative similarly strikes me as very similar to monad comprehensions.

Answer 1

No. You can write bind with monad comprehensions:

m >>= f == [ b | a <- m, b <- f a ]