Is any component of Sagemaker still involved in production inferencing

Are any components of the SageMaker Engine involved still during production ( inference) , I had a discussion with someone saying Sagemaker has nothing to do with production anymore because everything is deployed , containerized and executed somewhere in a VM or cluster., I think there are some instances the main engine is working in production and please correct me if i am wrong 1. Billing records - have to be generated -so something is connected to the engine ? does the model help here or is it just pure EC2 etc compute measured ? Who manages the endpoints ( server less, and asynch) ? if we need to use Sage maker monitor this is also alive during production. Question is What main components are active in production outside model or does mode communicate to engine still? As an addendum someone mentioned to me if it is deployed to EC2 or K8s yes it is disconnected from the main services in production you CANNOT use Model monitor in this scenario? --- but if you deploy to SAGEMAKER INFERENCE? you can use model monitor in prod. You deploy to an endpoint for real time... What is SageMAker Inference - as a concept or as some physical thing because there are 71 options in Sagemaker inference ?