I was about to vote against this proposal on Wikidata, but I will not, based on the principle that those who say it's impossible should not interrupt those who are doing it.
But I will give my opinion here. As you know I'm very much supporting the Wikidata-WikiTree connection. But, seems to me, there is work enough in matching well-defined entities, namely people, comparing what is asserted on both sides and try to improve either or both, based on alerts triggered by discrepancies (BTW, I would prefer all those so-called "errors" 5xx to be called simply "alerts").
But matching such fuzzy and elusive entities as Wikitree categories is conceptually tricky. One vote against this proposal on Wikidata is based on this conceptual mismatch risk, and I totally agree. Categories are conceptually a mess both in Wikidata and Wikitree, and mapping two messy schemes is just a headache.
Moreover, categories on both sides are likely to change any time. And they will. How will you be alerted of the changes, and if yes, how will you deal with the change? Mapping category schemes is already a headache when both sides are stable and conceptually sound. When neither of those premises is achieved, it looks to me as a non-starter.
So ... go ahead and good luck, but count me out .
EDITED : I would support mapping of other entities than people, e.g. locations, if those were represented in WikiTree model as entities and not as categories.