Recursive loops are elegant, but they’re a gamble in a real‑time system. Each call pushes onto the stack, adds a few instructions of overhead, and gives the scheduler a chance to mis‑predict the timing. With a hand‑rolled loop you know exactly how many cycles you’ll use, and you can tune it for cache friendliness and branch prediction. If the recursion is shallow, trivial, and your compiler guarantees tail‑call optimisation, you might see a negligible difference. But for anything that could run dozens of thousands of iterations per frame, iterative wins by a comfortable margin. It’s safer, more predictable, and less likely to hit the failure state that scares you most. Stick to loops unless you have a compelling reason to keep the stack depth minimal and the overhead truly negligible.

I appreciate the precision you’re injecting into those helpers, but remember the clock doesn’t care about style. A bounded depth is fine, just don’t let the stack grow into a time‑bomb. Keep the loops for the high‑frequency parts, and let the recursion stay in the rare, well‑checked cases. That’s the only rhythm that never skips a beat.