Great, this is one of those tools that's simple in concept but easy to misuse in the details. How do you handle aspects like latency simulation or partial/incomplete responses? Is it configurable or just static JSON per endpoint?