Hi!
I do not think a separate Wikibase instance would be needed to provide the data for Wiktionary. I think this can and should be done on Wikidata. But as said by Milos and pointed out by Gerard, lexical
I am worried that having two different data sets within the same instance would be a problem for tools working with the data, and for humans too. And frankly, I don't see too much benefit - virtually all added value Wikidata has now is working with the assumption of the semantics of Wikidata values and properties. Everything that pertains to lexemes, forms, etc. will have to be built separately, so why do it within the same site and have all the mechanics act as a split brain? I would think having parallel instance of Wikibase would serve the same goal much better, while preserving all the benefits of using the Wikibase toolkit and basic data model. Ultimately, it's the same as having separate databases vs. having one huge database (or even one huge table) with columns marking virtual partitions - the former is much easier to handle if the sets are completely disjoint, as we'd have between Wikidata and Wiktionary, as far as I can see. Maybe I am missing some benefit joint structure would produce?