The first statement increments the value of "count"(which I guess is a new key made then and there?).

Yes. Perl will "autovivify" a new entry if none already exists. The is pretty cool stuff. In other languages I would have had to size and initialize the structure. In Perl, I can just do that as I go.

But that array is never used later?

Correct. My code produces exactly the same structure as your code, but in more understandable way (at least for me!). The "values" really aren't needed (and yes this is an @array). I just did that because your code did it. The "values" will always be the same as the number key and always repeated the same times as the frequency. That could easily be computed. So really all that is needed is a single dimensional hash instead of two dimensions (see my code later in the thread).

update:
I'm not sure about your understanding of sort... Trying to further clarify... $a and $b are two hash keys that sort chooses for us - these are some chosen set of number pairs. We don't have to be concerned with the algorithm that sort uses, we just have to tell it how to compare a and b.