One peice of tooling I've wanted for Vitess is for it to make recommendations about sharding keys based on the statistics it collects + query patterns. Seems li

One peice of tooling I've wanted for Vitess is for it to make recommendations about sharding keys based on the statistics it collects + query patterns. Seems like it's in the perfect location to collect the data for that and remove the mental overhead associated w sharding. If much of the implementation has to be rebuilt, is there a path for that in this implementation?