Now that 3.0-preview4 is out, I’ve re-run the same benchmark and a more realistic one. In this post, I’ll present the results.