you may disable the proxy service and if you have WinXP with an enable firewall you may have a good chance the Spark chooses IBB in-band file transfer. The transfer will take about 10 seconds to start but it should run fine.
Alex is the file transfer expert, for me it looks like IBB is used also for the first transfer looking at:
Could you post the complete packets from both exchanges per chance? It is not possible from the information you provided for me to discern what is going wrong with the second transfer. The error in the first transfer is the recipient of the file transfer notifying the sender that they cannot make a direct tcp link to them. The initiator then knows to begin the file transfer via IBB, the fall-back mechanism. So, it really isn’'t a failure of the file transfer its a failure in one of the steps of the negotiation process.
I can safely say I have not seen that before. user9 is confirming to user6 that they have successfully connected to them. It appears as though everything has negotiated correctly. Do you see any errors in your Spark log file? My next step would to probably look at ethereal for that port on user6 and see the tcp traffic that is occuring.