Do you have a source or benchmarks for the last bullet point?

I am skeptical that optimizations like that wouldn’t already be implemented by postgres.