> There was another interesting use case I wrote about a few years ago that showed how changing DISTINCT to GROUP BY – even though it carries the same semantics and produces the same results – can help SQL Server filter out duplicates earlier and have a serious impact on performance.

I recently learned this is also what Amazon recommends when querying Spectrum. [1]

[1] https://aws.amazon.com/blogs/big-data/10-best-practices-for-...