Your question is Isolating Bottlenecks Without Logs. Take a moment with it on the right.
Talk me through your thinking if you like. When you're confident, submit your answer and I'll grade it like a real screen (7/10 or better passes).
How do you isolate a performance bottleneck in a distributed system when there are no explicit error logs?