Maybe because the “cache” vdev is not available? Hard to say. (You have quite the complex pool: 4 x RAIDZ2 data vdevs + a two-mirror SLOG + a two-mirror dedup vdev + a two-mirror special vdev + an L2ARC cache + hot spares)
Do you have a backup of your important data?
Snapshots are a property of a dataset. If you destroy the dataset, you de facto lose all your snapshots. Only a “checkpoint” can restore a destroyed dataset. (Or a fancy recovery process, with a lower chance of success, however.)
Somewhat related, if the “checkpoint” feature was exposed to the GUI, it might nudge users into making it a habit to use it when doing migrations, destructive activities, major changes, or just “once in a while” safeguards.