...or don't use it.

Scheme likes to tout it because of tail call optimization. Common Lisp can do it, but doesn't really shove it as a first principle of "Common Lisping" (and CL does not dictate tail call optimization).

I don't use it (save when it's necessary). I just iterate using the supplied iterators (do, dolist, loop, whatever).

I can't speak to other functional languages.

CL's iteration primitives are more capable that those provided by most languages, which often makes for a very clean approach.

Recursion is sometimes the easiest way to reason about a problem though, so it's nice to have decent support and TCO.

> Recursion is sometimes the easiest way to reason about a problem though

Yes, it's one of the things that allow FP algorithms to be more declarative. It's difficult to seriously argue that recursion is fundamentally more complex-- the contrary argument is more obvious to me.

Everything that's left is the fact that it requires people to unlearn things, which is uncomfortable and requires effort, but unfamiliarity is not an indicator of complexity.

I take it as: mathematicians like recursion, so they put some recursion in your recursion...

There's a strong argument to be made that mathematicians tend to build the simplest abstractions conceptually, especially in comparison to computer scientists. Personally, when I stumble upon some maths, I pay close attention. It has paid off.