How does your code go from the repo to the users? How do you deal with connection draining? Do you make sure that requests using the old code aren't interrupted? Do you spin servers up and down based on load? Do you have websocket connections that can last for hours?