We actually have customers that have migrated over far earlier than that going from say a R3.xlarge to our smallest production cluster and seeing 2-3x performan

We actually have customers that have migrated over far earlier than that going from say a R3.xlarge to our smallest production cluster and seeing 2-3x performance gains. Comparing Citus to single node Postgres isn't really an apples to apples comparison. Part of our performance gains come from sharding tables under the covers, so even when doing painful sequential scans you're scanning less data. Another part is that we parallelize queries. Yes, for the exact same hardware you have a different price, but the price to performance customers have seen has been much better.