[1) I have DSL, not 56K. I have no idea where you got that idea.]
I think he misunderstood you, see 2).

[2) "Hits" I am using to describe how many items appear in a search window. I am not sure how you are using the term.]
I think he referred to the term 'query hits'; where it is indeed a good thing for child/normal nodes to have a low query throughput.

[3) While the horizon numbers in Limewire are not necessarily accurate,...]
*grin* they are nowhere never accurate, I think.

[...if I notice that the number of available hosts is small relative to the number of connections, then it is because two or three Gnucleus ultrapeers are connected. If I make an effort to weed them out, hosts increase and hits increase, and thus my chances of getting a download.]
Well, I usually have one to two (sometimes three) Gnucleus Ultrapeers, and I am very satisfied with my downloads/uploads.
Also I took a look at downloading (I mean downloading from me) clients and I found them to be relative broadly distributed, according to 'market shares'.

But I have to say that I haven't made any statistical worthy notes - it would be surely helpful if you can investigate a little further, as no Gnucleus user wants such a bug (if it is) to persist.

Well, a new gnucleus version also came out featuring a rewritten ultrapeer part. Eventually the mentioned behaviour is already solved.

If you think this persists you can also take a look to [gnucleus.com], there's also a forum.

Greetings...