mTree replication is great but if you delete a file on the source it's deleted on the destination once the next replication cycle runs.
Mtree replication creates its own snapshot and we do not use user created snapshot while replicating the data. If you have any snapshot on source it will be replicate to destination and you would need to have appropriate space available.
More information for your reference:
MTree replication leverages snapshots to ensure that the destination Data Domain system will always be a point-in-time image of the source Data Domain system. Snapshots are frozen images of the MTree at any given time. For each MTree replication context the system will auto-create (and auto-delete) snapshots of the contents of the Mtree. This snapshot is replicated to the destination DDR. At the completion of the replication, the system will auto-delete the old snapshots in the source and destination DDRs, and then will continue the cycle of creating, replicating, and deleting snapshots.
better use fastcopy to copy the entire mtree to another folder structure
e.g. filesys fastcopy source-mtree-path destination-mtree-path/day1
do this once a day for seven days. if data is lost then can fast copy back any of the dated subfolder
it will require additional space for daily changes as you stated.
As mentioned above mtree replication creates its own snapshots. If you want to keep data fora few more days you can create your own user created snapshots. We have a lot of customers that do just what you are suggesting. The thing to consider here is your schedule. Quite often I will see user created snapshot schedules where they take on in the morning and they are not backing up during the day and then they take another one later that day just before they start the nightly backups. This creates an additional snapshot that is not needed because it is the same as the on taken earlier in the morning. As for space it will take more space. The additional space depends on your deltas between the the various snapshots and the duration of time that you keep them. If you have a high change rate the space you will consume will be greater than if you have a lower change rate. It is difficult to give a precise answer here because the answer will depends on so many different variables.
One other option to this issue is if you are using an application that uses ddboost and it supports mfr replication you can have different policies for retention of data between the source and target Data Domains.