The title is somewhat misleading: it is not the core Wikidata database which is splitting, it is just the WikiData Query Service (WDQS) which enables users to run SPARQL queries over its data.
WDQS is already very difficult to use – many queries run extremely slowly or time out completely, and working out how to write a query in such a way that it performs well requires a high level of expertise, both in SPARQL in general, and in the details of Wikidata's implementation of it in particular.
