There's a lot of excitement around the concept of federation in Wikibase. Generally speaking, federation means sharing information between Wikibase instances in order to broaden the base of available data.
However, federation takes many forms, and misconceptions or presumptions about the term can lead to confusion. Below we lay out different kinds of federation and explain what's currently available.
The federated query is a property of SPARQL that allows your search to include multiple endpoints.
Federated SPARQL querying works in Wikibase. You can see a partial, example list of federation endpoints used by Wikidata here.
Finally, there's an interesting blog post that demonstrates some federated querying in the context of the Wikibase registry.
Instead of creating a set of properties from scratch on a new instance of Wikibase, it will eventually be possible to configure a new Wikibase instance to obtain a robust, living and constantly updated property set from Wikidata.
Federating with Wikidata's properties doesn't amount to mirroring Wikidata. Though some want to do just that in order to overcome rate limits, and dumps are produced regularly, such an endeavor is by no means an easy task and likely won't accomplish what you're aiming for.
Two-way federation of data between Wikibase instances is not yet implemented and probably a long way off. The aforementioned Federated Properties project is the first step of a longer journey in this direction.