You are talking about hits, I am talking about bandwidth, if your client
sent it's requests in a bunch then bandwidth is the determining factor
not request-response RTT.
If you are referring to HTTP 1.1 persistent connections and pipelining,
then it would be better to use the terminology of RFC2616 so that we all
know what we are talking about.
Using pipelining does eliminate much of the latency effect, but you
still have the header overhead for each separate piece. A site delivered
in many fragments will require more downloaded data to deliver the
content than it would in fewer fragments. It also requires more HTML to
refer to many image fragments than to refer to one, so there is more
content to download for the same visual effect.