This has probably been my worst experience in system design interviews. And I've made many of them, both as a candidate and an interviewer. A video streaming system is a quite standard design question, and one where one can zoom in with many follow-ups (what if a system goes down? what if the user uploads on a mobile connection? how would you manage transcoding? how to manage redundancy/availability?). Unfortunately, my understanding is that Datadog interviewers don't expect to come up with many follow-ups. Adapting an existing design to fit new requirements is a minus point on their notes. So I did my interview as I'm used to (designing a rough MVP that could solve the basic problem and then expecting follow-up questions to extend it), I eventually leveraged the follow-up questions to come up with a solution that matches almost 1-1 those proposed on some system design interview books, but yet they rejected me saying that "the system had to be adapted to fit new requirements". Well, isn't that the whole point of a system design interview? Coming up with a first design that solved the problem and then scaling it up and refining it later? Isn't it what happens in real-world engineering, or are engineers at Datadog are expected to come up with the perfect solution that requires no follow-up adjustments in 45 minutes? I feel like Datadog should really adjust their expectations from system deign interviews, or they'll end up pushing away more and more valuable candidates.