You can speed up the loads, using the “QVD Layer”. QVD (QlikView Data), is a file sitting in file system, it has 1 table in it, and it is already compressed.
You would query Hortonworks in chunks, and create bunch of QVD files in parallel. You can have daily QVDs, or monthly QVDs or yearly QVDs depending on the use case,
Then you can load these QVDs to a QlikView app very quickly (seconds).
Take a look at this article, and the attached white paper.
also, try increasing the buffer size on the DSN.