Caterpillar

Closures are currently boxed (i.e. allocated on the heap), so they can be represented by a single value on the stack. I don't think there's much to do about that, while the language is still untyped. With a more sophisticated type system, it will be possible to have closures live on the stack, with no extra heap allocations necessary.

And there are other sharp edges. A closure is consumed (its heap allocation is removed) when evaluated by something like if. But nothing prevents you from copying the value that represents the closure and try to use that again. Or make up a completely new value that then might or might not refer to a closure.

Right now, there's not much point in trying to fix stuff like that, I think. Things will get better, as the language develops. It's a process.