Evaluate a performance test with mixed request outcomes.

Fictional case: a search API is tested after a separate warm-up using an open arrival rate of 200 requests per second for exactly 600 seconds. The generator sent all 120,000 scheduled requests without local drops; all completed within the measured interval. The agreed workload is 80% ordinary searches and 20% complex searches against a production-sized synthetic index. Of 120,000 requests, 119,400 returned correct successful responses and 600 returned errors. For successful requests, measured p50 latency is 180 ms, p95 is 480 ms and p99 is 1,400 ms; failed requests are separately reported and have p95 40 ms. The acceptance criteria for this workload are: at least 199 correct successful responses per second, request error rate no greater than 0.5%, and successful-request p95 no greater than 500 ms and p99 no greater than 1,000 ms. Generator telemetry shows capacity headroom. Do not assume results for a different request mix, data size or longer duration.