I can’t, off the top of my head, come up with how you would do this with a smart rule. But: by default there is a smart rule called “Filter Duplicates” - it will show you all duplicates (don’t run it - it would move your duplicates to trash). In the Data menu there is a function Convert/Duplicates to Replicates. So that should deal with 1-3 on your list (mark those files listed by the rule, and then select the aforementioned function in the menu).
Be aware, though, that you cannot replicate across databases; you can only duplicate across databases.
Without scripting* I couldn’t automate your points 4 and 5; if you don’t have thousands of replicates, the easiest way would be to duplicate the “Filter Duplicates” smart rule, change “Duplicates” to “Replicates” in the rule, add “location” to the columns in the View menu, and then simply go through the list manually, marking those instances which are to be deleted.
* even scripting would be relatively complex, because you aren’t looking to remove all duplicate instances of replicants, but only those multiples in the same group. @pete31 you haven’t got a script of this kind hidden away already?