@J_J_Sloan Thanks for taking the time to run and share these benchmarks.
If you happen to have it handy, it might help readers interpret the results even better if a bit more context around the test environment were included. Things like the VM configuration (vCPU count, RAM), underlying storage type (NVMe vs SATA, host filesystem), ZFS pool layout and recordsize, and whether any non-default mount or tuning options were used can make a meaningful difference in how these numbers are read.
Totally optional, of course! But this has peaked my interest. So even a short summary would add a lot of clarity for anyone trying to compare this to their own setups or reproduce similar tests.
Either way, thanks again for posting the graphs and results!