Yeah, it was more that I felt like I had fucked up the explanation. I didn’t want people to think we were using the tx/header hashes as hash table keys. Really it wasn’t important enough to worry about. But yeah, the nonoptimal bucket counts bugs me.