FS metrics without random IO benchmark are near meaningless, sequential read is best case for basically every file system and it's essentially "how fast you can get things from S3" in this case
It is all part of ZFS architecture with two tiers:
- Special vdev (SSD): All metadata + small blocks (configurable threshold, typically <128KB)
- Object storage: Bulk data only
If the workload is randomized 4K small data blocks - that's SSD latency, not S3 latency.