Thanks! I unfortunately wasn't really scientifically disciplined enough as I ran pgbench running through configurations, so I didn't include my notes or numbers but hopefully someday I can come back to this with a fresh wad of cash to spend on dedicated hardware fit for purpose and put some of this to rest!

One of the things about this kind of benchmarking/testing is that it really requires multiple days worth of evaluation (index bloat, disk fragmentation, etc) and is really hard to get right... I'm afraid I'd add more noise than signal.

Some great resources out there though, like Percona's post on MySQL:

https://www.percona.com/blog/mysql-zfs-performance-update/

Even in there though, the comments are full of reasonable objections, with some really sage advice in terms of fragmentation and really measuring whether ZFS can coalesce your writes. It's a hard topic!