Your metric is p99 latency. You cannot t-test a quantile
Test selectionHardProblem. The metric of record for the encoder test is p99 end-to-end latency, because the product is ruined by the worst moments and the mean hides them. A colleague computed p99 in each arm, got 96ms against 91ms, and ran a t-test on the two numbers. What is wrong, and what would you do?
Before you reveal: say your answer out loud, as if you were in the real interview — get your reasoning across clearly first. There is no single correct answer: reading what the interviewer is really after and defending your own thinking is what makes an answer strong.