41. How would you debug a production issue you cannot reproduce locally?
MediumTesting & engineering practice
Technical
Show model answer
Limit impact first (roll back, disable a feature flag), then gather evidence: logs, metrics, traces, error reports and recent deploys. Form a hypothesis, compare a failing request with a working one, and narrow the scope (one region, user type or version).
Add targeted logging if needed, fix with a regression test, and write a short blameless post-mortem.