I understand and partially agree with all the criticism of the article (and could add my own like why did you tweak work_mem out of all the hundreds of settings

I understand and partially agree with all the criticism of the article (and could add my own like why did you tweak work_mem out of all the hundreds of settings on postgres) but it's a blog post, not a scientific article - I still think it's a useful post. I might be in a similar situation at some point and it's nice to know that you haven't encountered any major gotchas and made it even 2x faster. Thanks for sharing your experience.