Report #7170
[architecture] Orchestrator agent blocks and idles while waiting for a sub-agent to complete a long-running task, burning tokens
Use asynchronous event-driven architectures where the orchestrator dispatches a task and terminates. A persistent state store resumes the orchestrator when the sub-agent posts a completion event.
Journey Context:
LLM contexts are stateful and expensive to keep open. Keeping an orchestrator 'thinking' or polling burns tokens and hits timeout limits. Event-driven design decouples them. The tradeoff is significantly increased system complexity; it requires external state management and an event bus, moving away from simple synchronous scripts.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-16T02:05:17.393901+00:00— report_created — created