Is there documentation on how to architect runpod serverless?

Wondering if theres Do's / Dont's of integrating runpod serverless into a larger architecture. I assume its not as snappy as lambda so I'd need to plan more aggressively around warm / cold starts? Also is RunPod serverless ready for prod deployments, or is it more of a "use at your own risk" service?