(Author) Thanks for the feedback! Yes, there's still a network load balancer involved — we debated whether we should include the last section of the post but de

(Author) Thanks for the feedback! Yes, there's still a network load balancer involved — we debated whether we should include the last section of the post but decided it's still worth writing out. The idea is that the load balancer doesn't need to directly connect to the running machines / containers, but instead just a much smaller set of `modal-http` service pods that translate HTTP to function calls, which lets us deploy and service big apps a lot faster. :) Writing a good network load balancer is very hard and we don't intend to suggest that it's our forte. We very much defer to all of the existing systems literature on this (and a couple papers like Maglev are in the post.)