This is more correct than allowing the two checkpoints to get out of
sync by hours; however, it's a departure from v0.11 behavior.
Signed-off-by: Zygo Blaxell <bees@furryterror.org>
[state]
+[state.point]
+
+ # Defer crawl checkpoint writes until the hash table writeback has
+ # flushed all extents to disk that were modified since the checkpoint update.
+ defer = yes
+
# The [scan.subvol] section controls legacy subvol scanners (scan modes 0-3).
[scan.subvol]