How did you track request durations, hanging requests, and other issues if distributed tracing was not available?