Why it is a question
In Devezer et al.’s model (LIT-988, read in full in NOTE-792), the consensus changes whenever a proposed model wins a single comparison, and scientists hold no beliefs. There a population of boundary testers reproduces almost perfectly while rarely holding the true model, and replication cannot change the transition matrix (THEORY-295). The authors call the single-test rule unrealistic and say it may exaggerate the role of mavericks. NOTE-792 lists the same question as open. Until 2026-10-19 it was THEORY-295’s promote_when. THEORY-295’s body disclaimed it, so the clause moved here when the title was narrowed to the model actually run.
What would answer it
A model in which the consensus has memory, because a challenger must win repeatedly or because evidence accumulates, and in which scientists can learn. It would answer yes if a population whose proposals never put the true model against the consensus still holds a false consensus with high reproducibility, while a population that keeps the model space connected does not. It would answer no if memory restores the coupling. Or a meta-scientific study of a real field whose replications succeed at a high rate while independent evidence shows its standing model to be wrong, traced to which alternatives were tested. Further runs of the memoryless 14-model framework cannot settle it, because its result is already proven there.