You can see the whole back-and-forth here: https://github.com/fivetran/benchmark/issues/1 https://github.com/fivetran/benchmark/issues/1 The tuning was pretty

You can see the whole back-and-forth here: https://github.com/fivetran/benchmark/issues/1 https://github.com/fivetran/benchmark/issues/1 The tuning was pretty aggressive, some of the sort keys were a little unrealistic for a real-world scenario where you're doing a lot of different queries.