The case
Two pieces of evidence support H, conditionally independent, prior odds 1:1, Bayes factor 10 for E₁ in both worlds. E₂ has Bayes factor 20 in world A and 1.1 in world B. E₁ is found unreliable and discounted. Posterior odds are 20:1 in A and 1.1:1 in B. “The unweighted graph has the same topology in both cases.”
What it can show
That which supports exist does not fix how a conclusion should move when one fails. A12’s “independent support survives invalidation” is the same structure seen from the review side: losing a ground need not defeat the conclusion, and how much it matters depends on what the other ground carries.
What it cannot show
A6: “That graph-based memory is intrinsically insufficient. An attributed graph could include likelihoods, dependence models, assumptions, and other relevant information.”