Hey HN!<p>We've just released this blog post exploring some of the changes we made over the past few years that helped us reduce memory usage by as much as 50% and improve throughput towards near-linear scalability.<p>We'd love to get your thoughts.<p>Cheers!