and should we really be tracking the gender of each genus as part of our data exchange model?

Good question. The Catalog of Fishes does, and I have found it to be EXTREMELY useful. The other parameter necessary for algorithmic gender-matching of species epithets to genus names is whether the species epithet is proposed as a noun, or as an adjective (not always obvious or self-evident). Only adjectives change to match the gender of the genus -- nouns preserve original orthography. At least that's how it works in ICZN-land.

To make this more relevant: If we chose (at least for zoology) to equate our nomenclatural objects with epithets rather than binomials, how do we expect applications to use these objects?

Appilcations would use them only in the context of a usage instance (wherein the rank, parent name(s), and orthography are indicated as attributes of the usage instance). As stand-alone objects, they would default to their original usage instance (i.e., original genus, original orthography, original rank). Authorship (and certain other Code-governed properties) would be preserved as attributes of the Name object, and the rest would be stacked on the usage instance. This fits in nicely with Jessie's campaign within TCS to encourage using names only in the context of a "SEC" (concept definition usage) -- and not as stand-alone "nominal concepts" -- whenever possible.

Are there any special use cases that we need to consider?

I'm not entirely clear on what you mean by "use cases", but autonyms/nominotypical names (i.e., same epithet, same rank-group, different rank) certainly need to be dealt with as "special case" -- but I think that is true regardless of whether coming at it from the botanical or zoological perspective.

Rich