the btrfs replication works on the concept of copying a subvolume, like a snapshot to another machine. However, as the replication continues to operate over time, additional snapshots will be created. However, btrfs snapshots are very space efficient (managing the deltas between an earlier and a later snapshot), unless you have major changes on the data you’re replicating all the time.
@phillxnet can probably provide a better/detailed answer to your concerns.
For your second question:
Here is a wiki-fied post from a long time ago, so it’s not accurate in all points anymore:
See the sections for pool, pool mount point, share and share mount point.
One is the subvolume with the underlying data, the other is the mount point (exposing the data to be used by things like the WebUI, Samba, Rockon’s, etc.).