You kick off a load test on your OpenAI integration. By the time it fails at 100,000 requests, you've spent $3,000 on tokens. The failure wasn't in your application logic.
Source: [Dev.to](https://dev.to/forgeflows/why-llm-load-tests-are-costing-you-thousands-2hei)