Runaway Token Consumption Cost Spike

Track gen_ai_client_token_usage and langchain_llm_cost per request and per hour. Alert on token usage exceeding 5000 tokens per request or hourly costs exceeding budget thresholds. Monitor langchain_agent_intermediate_steps for excessive iteration counts.

Set hard limits on max tokens per request and max agent iterations. Implement cost guardrails with automatic circuit breakers when hourly spending exceeds thresholds. Track cost per user/session to identify high-cost patterns. Add token usage forecasting based on langchain_llm_cost trends.