I wonder if it's worth taking a page out of the MongoDB book and enabling these kinds of benchmark altering settings by default. We certainly selected clickhouse over tailscale internally because of major performance differences in our internal testing that might have gone the other way had we "known better".