Quick clarification:
Am 15.09.2016 um 17:40 schrieb Jan Berkel:
The pdf mentions 4 new entity types: Lexeme, Statement, Form, Embedded (?).
"Embedded" isn't a separate type, it refers to the fact that Senses and Forms are stored on the same page as "their" Lexeme. "Statement" isn't an entity, I assume you meant to write "Sense".
Curious, was the existing data model not flexible enough?
It was not expressive enough, no; it would be possible to use items to model lexemes, but it would be very annoying and complicated. You would need separate items for each form and sense, and need to keep track of them for deletion, undeletion, etc.
Will these new entities be restricted to the usage in a lexicographical context, i.e. Wiktionary?
It will also be accessible from Wikipedia and other wikis.
How will they fit into the existing data model, will there be links from existing Wikidata items to the new entities? (i.e. how will Wikidata benefit from the new data?)
Yes, there will be cross-linking.
I imagine in an integrated Wikidata/Wiktionary world "content" and code lives in various places, and we'll have a range of automated processes to copy things back and forth, and to automatically create new entries derived from existing ones?
It would be transcluded and generated, like with templates and and Lua. Not so much copied, with bots.