I can definitely see an argument for going pure FP and for using simpler languages. The trade-off though is that it puts you further from the other languages they will learn later, both syntactically and conceptually. Lots of places have used Scheme/Racket, but they have the dynamic typing issue. (Note that these days I consider that as a late-feedback issue as well because students aren’t told about an error until it executes and trips it.) The statically typed options are SML, O’Caml, and Haskell. Those could work assume one doesn’t go too far down the rabbit hole. How much do CS1 students really need to see Monads and Typeclasses?

As with all things in CS, there are tradeoffs. There is no right answer. Which answer is best for you depends a lot on what you value. I’m willing to accept a bit of complexity to form a good base for future learning in other languages. Others could feel differently.