> given any system of sufficient complexity,

That's my question. It seems that adding "self reference" to the system makes it more complicated than it would be otherwise. So it's not just a system with enough complexity to let you do arithmetic. You need a system with enough complexity to also do self-reference. And I would say and think that self-reference is complicated.

Do we need self-reference in any other areas of mathematics, than Godel's and related proofs?