You're right, I misspoke. H-W deals with allele frequencies and how they change over time leading to specific ratios (equilibrium). It also shows that a variant can never be lost from a population once it's there.
One of the assumptions of H-W is that of an infinite population size, so that genetic drift does not occur.
Among others like completely random mating, non-overlapping generations, no mutation, no immigration/emigration, and no natural selection (although you can add a selection coefficient to H-W and get a reasonable result that still fits with the original equations).
In a particular model. Models are not reality, and are only valid to the extent to which they have been tested against reality.
Right. That's why I'm not a theoretician. I prefer to use models that at least attempt to reflect something like reality. H-W doesn't reflect reality, Orr comes a bit closer but he's dealing with something completely different than the basic pop-gen equations.