I've just lost data following the selected answer to this question, so I'm posting this to clarify. Please correct me if I'm wrong.
IMO, LVM terminology is backwards. For reference redhat's documentation to LVM snapshot
When LVM says "drop a snapshot", it means merging the changes from the COW volume into the origin volume.
When LVM says "merge a snapshot", it means reverting the origin volume to the state when the snapshot was made, effectively dropping the changes stored in the COW volume.
I find this VERY counterintuitive, lost 43h of data because of that. Don't do the same mistake I did !
EDIT: OK now I understand my mistake. From the documentation:
When you modify the original volume (the origin) after you take a snapshot, the snapshot feature makes a copy of the modified data area as it was prior to the change so that it can reconstruct the state of the volume. When you create a snapshot, full read and write access to the origin stays possible.
I wrongly assumed that LVM operated by freezing the origin partition and mounting the snapshot as an overlay to store the modifications, but this sentence makes it clear it does not work that way.