Judging what is and is not evidence is an important task that has been completely ignored here.

In the most literal, unbounded application of Bayesian induction, anything within the past light cone of what is being considered counts as "evidence". Clearly, an immense majority of it is all but completely independent of most propositions, but it is still evidence, however slight.

Having cleared up that everything is evidence, determining the weight to give any particular piece of evidence is left as an exercise for the reader.