As we all initially expected it to do. 
I guess the original intent of “Synch Dataset” was to do exactly that. But as you probably saw in the RDL(after pressing Sync Dataset), that process is unable to simple add the new additions to the RDD’s output. It has to breakdown the RDD into individual tables(which appear as separate Datasets in the RDL), renaming fields where conflicts exist. And often fails at this.