I am assuming you are creating connections based on a process that simply listens to connection attempts. The TCP/IP functions don't provide a means to detect a broken connection, so you basically have to roll your own, or try to hook into the operating system.
In LabVIEW you can try having a simple monitoring loop that performs a TCP Read operation on each connection. This loop doesn't have to do anything with the data. The intent is to see if you get an error out of the TCP Read. In the event of a closed connection you should get error code 66. You'll need to experiment with a reasonable timeout that is long enough to detect a lost connection, but not too long that it takes a long time to check an individual connection. Note that I'm not sure if the TCP Read returns code 66 or a "timeout" code.
You can read more in this thread:
http://forums.ni.com/ni/board/message?board.id=170&message.id=297431
