https://phys.org/news/2018-06-evidence-scientists-decades.html

A common way scientists gather evidence is to make a prediction about something and see if they're correct. The problem occurs when the prediction is right but the theory they use to make it is wrong. Predictions that seem particularly risky but turn out to be true look like very strong evidence, as Karl Popper and other philosophers of science have often stressed. But history shows us that even very strong evidence can be misleading.

In 1811, Johann Friedrich Meckel successfully predicted that human embryos would have gill slits. This risky prediction seemed to provide very strong evidence for his theory that humans, as the "most perfect" organisms, develop via stages corresponding to each of the "less perfect" species (fish, amphibians, reptiles and so on).