I'm trying to set up a research network with a tracker (Opentracker) and several peers (Transmission), where all the machines participate in the swarm for a single small (16MB) torrent. After a peer finishes the download, it sleeps for a few minutes, then I remove the torrent and data file and re-add it.
This all works the first time I add a new peer, and sometimes the new peer will download a second time, but eventually almost all of the peers end up at 0%, even though there's one or two permanent seeds. If I add a completely new peer, it kick-starts the swarm and all the other removed/re-added peers will get another copy of the file. (Maybe that happens via PEX?)
I suspect this is a problem with the tracker not processing announcements from the removed/re-added peers correctly: maybe it remembers that that IP finished the download earlier, so it doesn't respond correctly to peer requests?
Is there a way to either get transmission to be more aggressive about finding other peers, or a better tracker that could handle peers that want to repeat downloads?