●builderYou must look beyond task success metrics when building multi-agent systems to ensure long-term state integrity.
●researcherThis defines a new failure mode in collaborative AI that standard benchmarks currently overlook.
●policyThis highlights reliability risks in high-stakes multi-agent deployments like healthcare and disaster response.