By monitoring responses, Airlock Gateway can detect problems with back-end hosts, and decide to exclude individual servers from receiving further traffic. By allowing some traffic to a "bad" server, Airlock Gateway can also detect once such a server has come back, mark it as good, and start sending regular traffic to it once again. Health checks can either be in-band (using regular, end-user requests) or out-of-band (requests for a pre-defined page). back-end failure or success can be detected by analyzing HTTP status code and the response content.