To be precise, I think that it does not “both fail and finish”.

You get the success message a little late, and some job control has decided earlier that it has failed, when in truth it just took longer than expected. The ‘fail’ message comes from this job control, while the ‘success’ message comes from ZFS itself.