Request-Response - growing memory -> out-of-memory exception


I got an "out of memory exception" with my productive system. For testing, I used the Request-Response example from the manual. Every port was running on an own host instance and I was using the TCPClient and the TCPServer as well as a tool (mousemover) which clicks connect, send and disconnect on the TCPClient.
This test was not very fast but it was running for ~5 days and eventually I also got an out of memory exception (as with my productive system). I tested it on BizTalk 2006 and BizTalk 2009.
In more detail:
At start every the host instance used ~10MB.
After several days the receive host was using ~500MB and the sendhost ~250MB.
The memory was not going down again, not even after the TCPCLient was stopped.
When I then restarted the TCPClient, the memory usage of the receivehost was growing up further up to 1,4GB and eventually the host instance crashed (i.e., it threw an out-of-memory exception and then it restarted automatically).
Other adapters like the SQL Adapter start with 10MB, then they may use up to 100 MB, but then they go down again to 10 MB when I stop sending messages.
Could it be that either the garbage collector does not work properly or that there is a problem with the request response funcionality? (If I use a one-way receive port, then there is no memory problem. That works pretty good! The problem only occurs with a request-receive port)
So my question is, could it be that there is a bug or is there anything I can do?
Thanks for your help!

file attachments