What major scalability blind spots do performance engineers invite when they rely entirely on lightweight, protocol-level load testing tools like k6, Locust, or Apache JMeter rather than heavier, real-browser simulation platforms? Furthermore, writing execution scripts in JavaScript or Python allows seamless integration directly into CI/CD deployment pipelines, but completely strips away frontend processing metrics, JavaScript execution overhead, and Document Object Model rendering bottlenecks. How do you justify optimizing backend microservices for millions of concurrent hits when a client's real-world browser environment crashes under standard user traffic?