We have the Quickfix engine operating in Acceptor mode. We frequently see scenarios where we disconnect the clients because we "Timed out waiting for heartbeat". But I am able to confirm from the network packet capture that the client heartbeats as expected and in fact promptly responds back to the test request (which was sent before the disconnect) with a heartbeat. I did search for this online and noticed a few forums were this scenario is being discussed, but no resolutions. We have seen similar scenarios in other applications when we don't set a proper socket read timeout and I am wondering if it is the same case here.