Discarding accounting is not my experience with most NASes I have worked with. They will keep on retrying until they get a response for STARTS and especially STOPS and not give up after 3 attempts unless you explicitly configured it on the NAS to discard STOPS. If you have configured the NAS to discard accounting then sure if that is what you wanted. But a NAS normally expects a response to EVERY message it sends and when it doesn't get a response it will keep on retrying. Such as if it has a round-robin configuration with multiple radius server endpoints it would typically give up on one, and then try another. But ultimately it is entirely the control of the NAS and the Radius server no matter what variant it is should always respond to any requests it gets unless the shared secret is wrong and it has been configured not to reply. To have a Radius server not respond to valid accounting messages that were initiated by the NAS is a terrible idea. The few use cases where you don't want a response to a valid request is typically with proxied requests where you are sending the messages to multiple servers with enough redundancy so they will reconcile at some point. So in short I wouldn't do it.