> there needs to be a long period of existing as a proto-science where people aren't doing experiments and are just observing and describing.
> before darwin
There was an understanding of natural selection even back to antiquity. How could there not be? Did people not tame the animals and plants? These are experiments, and they saw the results.
There were great contributions to astronomy long before Kepler. There were many experiments that influenced the whole field. There was a lot of important chemistry that happened long before Lavoisier (conservation of mass) and Dalton (atomic model).
The proto-sciences are nothing to scoff at. They aren't useless and they weren't ill-founded. They were just... noisy (and science is naturally a noisy process, so I mean *NOISY*). There's nothing inheriently wrong with that. The only thing wrong is not recognizing the noise and placing unfounded confidence in results. That famous conversation between Dyson and Fermi discussing von Neumann's elephant wasn't saying that Dyson didn't do hard work or that the work he did had no utility, it was that you can't place confidence in a model derived from empirical results without a strong underlying theory. You'd never get to that if you only observed because you'd only end up making the same error Dyson did.
Science, in its nature, is not about answers, it is about confidence in a model that approximates answers. These two things look identical but truth is unobtainable, there is always an epsilon bound. So it is about that epsilon! Your confidence! So experiments that don't yield high confidence results aren't useless, but they are rather just the beginning. They give direction to explore. Because hey, if I'm looking for elephants I'd rather start looking where someone says they saw a big crazy monster than randomly pick a point on the globe. But I'm also not going to claim elephants exist just because I heard someone talking about something vaguely matching the description. And this is naturally how it works. We're exploring into the unknown. You gotta follow hunches and rumors, because it is better than nothing. But you won't get anywhere from observation alone. Not to mention that it is incredibly easy to be deceived by your observations. You will find this story ring true countless times in the history of science. But better models always prevail because we challenge the status quo and we take risks. But the nature of it is that it is risky (noisy). There's nothing wrong with that. You just have to admit it.