As of right now, we send 4 to 8 GET requests to the S3 storage on server startup in order to download the latest server, or to verify if it has to be updated. In order to limit billing and reduce time the server takes to setup, we could try to limit the number of query when loading the models.