From the Racket docs: This evaluation behavior is sometimes called tail-call optimization, but it’s not merely an “optimization” in Racket; it’s a guarantee ab

From the Racket docs: This evaluation behavior is sometimes called tail-call optimization, but it’s not merely an “optimization” in Racket; it’s a guarantee about the way the code will run. More precisely, an expression in tail position with respect to another expression does not take extra computation space over the other expression. And you're right -- typed racket is still in its growing stages. There are a lot of guarantees you don't have.