If the service had been failfast the IdP wouldn't have started, so it's working exactly as intended. I imagine you are also ignoring the fact that each individual resolver has its own fail fast settings. -- Scott