How should Kubernetes resource requests and limits be implemented for production?

Measure real usage, set realistic requests and limits, monitor throttling and OOM events, use Vertical Pod Autoscaler recommendations carefully, and revisit settings as workloads change.