[BeeOnRope]

> Ok, I was just mistaken in my belief that integer overflow shenanigans was a major contributor to how a modern compiler optimized e.g. loops.

Yes, there is some impact but I think it's large, at least based on looking at a lot of assembly, and going over the typical examples of where it helps. In the cases it does help, it could make a big difference on a loop, let's say 2x the speed, but these aren't all that common.

> Right. I was considering the sandbox in the sense only of process security rather than program/type soundness.

Usually those two things end up tightly bound together: it's hard to impossible to enforce a sandbox if the user can escape the type system.