I believe there is also a good way to prove that some models work as we expect. I mean, we have now hard evidence that an given asteroid has properties X, Y, Z

I believe there is also a good way to prove that some models work as we expect. I mean, we have now hard evidence that an given asteroid has properties X, Y, Z - and we have mathematical (or ML) models that infer such properties from indirect measures. At least, we have evidence to validate that models or spot errors on them with ground truth.