Ah, to me “emergent behavior” has at least a partial connotation of some expectation on the part of the designer. Like, in robotics or AI you intentionally build a system with pretty simple individual pieces that display an “emergent behavior” when combined.
But that’s probably influenced by my field, because “emergent behavior” applies to a lot of natural phenomena too that aren’t clearly “designed”.
Here’s my perspective: In the same way you might design a reward scheme for an AI that leads them to take an absurd solution to locomotion, the designers of those physics systems designed them in a way that allowed for creative player solutions. Like you point out, it’s relatively easy to clamp down on emergent behavior if you care to, or your vision is different.
Emergent behavior is always edge case material, otherwise it’s not emergent. Flexible systems yield diverse results.
Perhaps this is the place to share it: I cowrote a proceedings paper for a simulation conference that talks about emergent behavior and metabehavior in videogames. It’s old and I’d definitely rewrite some parts, but I’ve committed to letting it sit as it exists now, warts and all. You might find it interesting.
I think this is true after the fact in retrospect, but not beforehand! For sufficiently complex systems, it’s very hard to prevent unless you have extensive experience in the niche you’re working in, or you’re building on something already existing and well-tested.
And I think that’s doubly true for earlier video games, which in most cases had engines that were built from the ground up.