Zfs has built-in encryption that can be enabled selectively per dataset, which is handy for some private data.
Also, as long as you avoid raidz, you can now add/remove disks to a pool, and attach/detach mirrors too. That’s much more flexible than it used to be.
Too bad raid1 from btrfs has not a zfs equivalent. You can to some extent replicate that via partitioning in zfs, but the simplicity of btrfs raid1 is great for uneven disks.
Has btrfs got its act together with large amounts of snapshots? It used to be a pain point.
Points 2 and 4 drive me nuts. For 2 in particular there must be some rationale, but the amount of times I’ve needed to see something behind the dialog, to have to cancel after having navigated the filesystem… Just bonkers.