Good catch, you actually found a bug. The predation (and dormancy) weights were being multiplied by scarcity/emotion modifiers, but in some early-return paths the original values were never restored. That meant the weights compounded every tick and could blow up exponentially into these absurd numbers.

That also explains the slowdown: once those values get extreme, a lot of the probability and normalization math gets very expensive.

It’s fixed now by restoring the original weights before every early return. Thanks for flagging it. That was a legit bug.