I heard a rumor that ZFS native encryption is somewhat abandoned by its maintainer following it having landed in the release, that it was one person and they moved on/lost interest. I don’t know if this is true or not.
I’m still using ZFS on LUKS2 for now, given the issues like this. (You also can’t specify multiple keys in ZFS native encryption.)
Sadly there have seemed to be a number of rough edges that have stuck around with ZFS native encryption, which is unfortunate given that the cross platform and remote replication advantages are fundamentally very compelling. Honestly it's kind of curious that polishing that experience hasn't been more of a priority, for commercial interests like iXsystems if no one else. Speaking of which, for people running their ZFS off a NAS the idea of just being able to offload all the encryption and FS computation to that is a compelling part of the value equation beyond sheer storage in principle too. Though that in turn (if one does it with iSCSI) runs into another seemingly neglected area of ZFS (zvols). Nothing is perfect.
What makes you say zvols are neglected? In fact the people that did most of the work porting ZFS to Linux, i.e. LLNL, use zvols for their HPC cluster and they still employ a number developers and the primary maintainer of ZoL. I also remember a quote from them stating that zvols were in fact more mature than vfs.
Since LUKS is exposed as block device like a hard drive, it's really no different than saying that they are just having issues with certain kind of hard drives.
I’m still using ZFS on LUKS2 for now, given the issues like this. (You also can’t specify multiple keys in ZFS native encryption.)