The main reason, here, was to avoid circular dependencies. This cluster manages an HA Harbor install. Rancher would then use it to bootstrap itself which obviously doesn't work. With multiple rancher servers, we needed something external that would service everyone without being dependent on a specific rancher server.