How would you debug an issue that only happens under production traffic and cannot be reproduced in staging?

Core Summary

Production-only issues are almost always due to data volume, traffic volume, or latency differences from staging. The key is to make production observable without breaking it: sample traces, shadow traffic to staging, add feature flags to disable the suspicious feature, or re-enable verbose logging temporarily. Never add load to staging trying to reproduce it; instead, bring production insight into your development environment.