BOUND sockets don't clear

Hi All.... I have a question that Informix Tech Support has not been able to handle yet. They have tried to tell me that this is expected behavior. Maybe one of you has seen this behavior. I recently had my production server stop accepting new connections although all of the user currently connected were having no problem. When I did a netstat -a, low and behold there were 32k worth of sockets with a status of BOUND. Coincidently, when doing a onstat -g nta, there was also aboyt 32k worth of threads with a name of unknown. By bouncing the server, all of these "hung" sockets cleared and we were able to continue. We have the default keepalive in sqlhosts and the kernel keepalive defaults to 2 hours. I can reproduce this at will from either a front-end pc or a dbaccess session on the server. For example: I first bounce the server to make sure that there are no "hung" sockets. I then start dbaccess and do a "connect" via the net (not shared memory). This gives me an "established" connection in netstat -a. I then exit from the dbaccess session. The netstat now shows me a BOUND socket. This does not clear. Evidently there is a max of 32k connections for a given port. I have tested this against 7.30.uc3, 7.31.uc3, and 9.20.uc1 (IDS.2000) and the behavior is consistant. I am on Solaris 2.6. Anyone seen this before?