Walk me through your approach when a production service experiences a sudden spike in latency but no explicit error logs are generated.
