Dear Maintainer,
snapshots of cached volumes seem to be seriously broken
(maybe it's fixed in the latest upstream version, I'm not sure).
To see the issue, try the following:
* create a cached volume, make a ext4 filesystem in it, mount it, and put some
contents there so the FS isn't empty;
* start some process making changes to the volume (like cyclically copying
files etc,
which imitates a busy filesystem);
* create a snapshot of the volume;
* try to check the snapshot with fsck.
With non-zero probability you will get filesystem errors on the snapshot.
(Note that if you do the same on a non-cached volume the check passes
as it should.) Trying to fix the errors automatically and then removing
the snapshot will most likely lead to the ruined filesystem on the volume.
I learned this the hard way when I tried to e2scrub my cached /home partition.
Granted, I have backups...