Think about how application-level health checking will work for the future. This will be in addition to #22 and should take care of application-specific health on each provider, allowing for an application to be healthy if it passes health checks and traffic is not routed to a bad instance.
