There are two kinds of circumstantial evidence that are useful for figuring out if D copied P. (1) Evidence suggesting D had access. (2) Evidence showing the degree of similarity between two works.
Evidence of access requires a reasonable opportunity (more than a bare possibility) to view or copy P's work. It cannot be inferred through mere speculation or conjecture. To do this, you might establish a particular chain of events between P's work and D's access to that work (i.e. tell a story). You might also establish that P's work is widely disseminated. For music, look at record sales, radio performances, sheet music sales. Additionally, subconscious copying has been accepted since it was first embraced in 1924 (memory as a trick is not an excuse). But this is more a feature of saying, "you had to have access because it was so popular, even if you don't realize you don't remember." I think this is something we can all buy. I know I say things all the time that are trademarked catchphrases because they've simply seeped into my thoughts...damn talented marketers.
Second, striking similarity can be so great that proof of access is presumed and need not be proven. The better the story you tell about access, the less you have to show striking similarity. Likewise, the more strikingly similar, the less you have to tell an access story. How this is handled varies by circuit, which I can get into if you'd really like. But generally, it’s not similarity, per se, that establishes access; rather, similarity of two works tends to prove access in light of the nature of the works, like the particular musical genre involved, or other circumstantial evidence of access. Here you might look for unexpected departures from the norm, or error. And like you said, lots of expert testimony is useful here - but why shouldn't it be? If I'm comparing two songs, I want an expert to tell me whether the two's notes are so similar that there had to be copying.