Excellent post. I don't think I'm ready to wield those sharp implements with quite so much flourish yet; the chance of lopping off my own limbs is currently too high.

However, there are some specific parts of your post I disagree with.

You have to shut that voice out. Ruthlessly. Because it has no way of knowing. That voice is designed to assess the motivations of members of an ancestral hunter-gather band. At best, it may have the ability to distinguish the correct murderer from between 2 and 100 possibilities -- 6 or 7 bits of inferential power on the absolute best of days. That may have worked in hunter-gatherer times, before more-closely-causally-linked physical evidence could hope to be evaluated. (Or maybe not -- but at least it got the genes passed on.)

Agree completely. Now apply it both ways. Specifically:

But the prior probabilities of such scenarios are low, even in general -- to say nothing of the case of Knox and Sollecito specifically, who, tabloid press to the contrary, are known to have had utterly benign dispositions prior to these events, and no reason to want Meredith Kercher dead.

This is extremely weak evidence. If you think that psychology doesn't really matter (and I agree with that), then it notmatters in both directions. There's no obvious motive, but that's very weak evidence that no motive existed. We just don't have enough evidence for what these people really think like, relative to the population at large. Re. motive, I'd definitely stick with the prior for murderous intentions among associates.

nor, while we're at it, do the 100 picograms (about one human cell's worth) of DNA from Meredith allegedly on the tip of a knife handled by Knox, found at Sollecito's apartment after the two were already suspects [count]

I was surprised at you dismissing the knife DNA evidence. The linked FOA page claims:

Low Copy Number (LCN) tests, like the one performed on the knife blade, are regarded by many experts as inherently unreliable, because it is not possible to prevent biological contamination at the level of picograms. Even in well-run labs, control samples regularly show up with DNA that theoretically should not be there.

The actual experts doing DNA-testing in the case evidently thought that it was significant evidence. Wikipedia has an article on LCN, though a lot of the citations are broken. According to the article, the method went through an extensive review, and was finally cleared for use in the UK justice system (and with a fairly definite statement at that: "The CPS has not seen anything to suggest that any current problems exist with LCN"). The situation in some other western countries appears to be rather different, though. I still think this is fairly strong evidence, though much less so than I did before doing some research on the general acceptance of this test method.

While not particularly flattering to the defendants (how would you like to be told that there's a 35% chance you're a murderer?)

Keep in mind this verdict uses (among other pieces of information) the existing jury verdict as empirical evidence (and of course in our epistemological position, that's the right thing to do); if our judgements had been based on just the primary evidence - i.e. without an existing guilty verdict by a jury - these probabilities would have been lower.