> A new model that is designed to fit existing observations that don't match our expectations based on the old model, without it correctly predicting new unexpe

> A new model that is designed to fit existing observations that don't match our expectations based on the old model, without it correctly predicting new unexpected observations, seems like a rather weak proposition to me. This description roughly applies to General Relativity; there are clearly merits to both approaches.