unbound VS bound

Here's an example of using LFresh for the simply-typed lambda calculus. https://github.com/sweirich/replib/blob/master/Unbound/Examples/STLC.hs Note that LFresh and FreshM are monads, so they will only guarantee fresh name generation while in the monad. contFreshM lets you "run" the monadic computation, so you'll only want to call it at the "top level" of your program.

First off, I am using Rust. I've begun by implementing an idea that is close in essence to bound, available in this playground. I believe that this is a good first step, but I am concerned with all the recursion leading to stack overflows and that analysing a term under a scope requires eliminating it, which in turn requires full tree traversals. There are ways to optimize the approach (such as caching variable occurrences in each scope node and skipping subtress when a variable I am looking for is not in it) but I wonder how far I can go with this approach and if there’s alternatives.

Has anyone applied this bound Imperative example to a larger language?