To add to above, how does parallelism work in Flink SQL? I found a stackoverflow link but there is no answer.