Again, if your system is "incorrect" having been stopped and snapshotted like that, it is also unsafe vs. power loss, something ZFS cannot save you from. Power loss events are vastly more common than poorly checkpointed database[1] events.

[1] FWIW: every database worth being called a "database" has some level of robust journaling with checkpoints internally. I honestly don't know what software you're talking about specifically except to say that you're likely using it wrong.