Dealing with multi-agent systems in production has been a headache lately, especially with tool-call storms crashing our orchestration layer. We’ve had to implement strict rate limiting just to keep things stable during high traffic multi-agent orchestration