For the past 4 or 5 days I can't reproduce the connection error in 2.9.8.2, 2.9.8, and even spent a day with Acquisition 0.861.

I did manage to make LW crash a few minutes after unchecking "disable ultrapeer" in the speed section of the preferences on OSX once, and logged 4 other crashes, but LW and Acq were consistently able to connect 3 Ultrapeers long enough to search for the same 6 "rare" .mpg's. (btw, LW won the searching contest well enough to find about 4-8 hosts for most, and completed d/ling 3 of them). Uploads were steady and all 5 slots stayed filled, but few completed: speeds were rarely above 1KB/s. Stats showed about equal up and down traffic of about 1GB each way in 24 hrs. I used the spinning Lime to time the space between the fairly narrow searches.

Firewall is open to 6346, NAT is mapped for TCP 6346 (I even added ANY 6346), and IP was forced except for today when I left it unchecked--didn't seem to make any difference in the stats. [trap_jaw, do I need to force IP if the firewall is open and the NAT mapped? seems redundant]. This is running OS 10.2.5 on a 700 MHz G3 laptop to a cable modem capped at 1 Mbit/s, sharing 220 files/ ~5GB. No one else on my NAT ran a gnutella client, but other Internet services were working quite consistently.

Saw lots of other problems (Could not move to library, 5 unknown crashes when LW left unattended), but no connection problems.

The two things I did that might be a bit unusual was to block any client returning a lot of spam (anyone want a list of 509 blacklisted IP's? they should be good for a few days before the DHCP's change. I think this is why LW had more success finding and retrieving "rare" files, so the hope of better queries looks to be happening) and to occasionally prune the Ultrapeers in the connections pane: I liked to see Acq .86's and LW 2.97 or better.

So--any suggestions on how I can reproduce the connection problem? I only have a day or two left before I give up gnutella (Lent is past) and go learn about HTML with the .mac homepage.

Cheers--sorry about the length.