In most cases it ends up just being extra cloned data. There's never really a reason for some networks to have different mutation settings than others, could prob use statics or something (statics bad). Might require adding more features to genetic-rs to allow simulations to contain arbitrary settings that crossover/division reproduction functions can access.