Great post!

A couple of thoughts:

1. It'd be interesting to see these benchmarks "under real-world adversity". For example, as data is ingested into the DW, as data shapes change, as concurrency of workloads varies.

2. It'd be great to measure "effort" to reach and maintain this performance.

DW vendors are trained on optimizing for static TPC benchmarks with concurrency of 1, which are useful, but can miss the boat on what users actually experience.

(BQ PM)