Ew. 
What are you, some kind of pig?
A ZFS incremental replication is exponentially faster than incremental rsync runs. There’s no need to populate a massive directory/file listing on both ends prior to running.
Replications also guarantee the same exact filesystem, down to the very blocks, permissions, and snapshots.
However, the main caveat is if you prune heavily on either side, and accidentally lose a “common” base snapshot. You’ll have to start all over again from a full replication.
You cannot. Once you commit to a source → destination replication process, it’s only one-way, and must remain so. Any and all changes to the destination must come from the source.
See these examples.