It's probably another of my misunderstandings, and I'm sorry to rejoin this debate, but I think I'm missing something.

Is it not the case in the swarming system of downloading from multiple hosts, that a file is created for each potential host? Each file holds the metadata description of the whole file to be downloaded and will keep track of which packets of data are missing in the downloads to date. Each file also has an imperative to attempt to fill all those identified holes. Hence, until one of the metadata records shows a complete download, I do not see why Limewire will not potentially request as many copies of individual data packets as there are hosts in the given grouping.