Since the protocol relies on passive timers to determine failure and take action, there is always the risk that a busy client will be slow in sending a keepalive response message back to the AppServer. Careful tuning of the timer values is necessary to avoid this risk of disconnection of healthy client connections. Note that your AppServer implementation details may add several seconds to the time it actually takes for the AppServer to detect the timeout.