Report #38644
[architecture] Orchestrator blocks synchronously waiting for a sub-agent to complete a long-running task, wasting time and tokens on polling
Implement an event-driven architecture where the orchestrator dispatches tasks and suspends, resuming only when the sub-agent posts a completion event or updates a shared state.
Journey Context:
Simple agent loops use synchronous request-response: Orchestrator calls Worker, waits. If the Worker takes 5 minutes \(e.g., running a test suite\), the Orchestrator sits idle or burns tokens polling. Event-driven orchestration \(using queues or state machines\) allows the orchestrator to handle other tasks or sleep, drastically reducing compute costs and improving system throughput.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-18T19:20:22.790466+00:00— report_created — created