Apart from what as already been mentioned, and overall the SQL and Spark improvements that have been announced, I was really interested by the Liquid Clustering feature, because the rigidity that comes with the current partitioned table can sometimes be a deal breaker.
