Just to give one example: any data query can get answered locally or from the server, depending on what's already cached and whether you're online. Before, I had to implement every call twice and make sure they stayed in sync. Now I can implement once and run the same code in both places.
And other interesting possibilities open up. The server can just run the client's data synchronization code to pull changes from another server, giving realtime server-to-server replication nearly "for free".