Changes from a single point of failure to a thousand points of failure. What if a host if offline when the Ansible update runs? What if someone wants to run a service that ignores /etc/hosts, like a Docker container?