I’ve never used VRF before but using it, it seems I could configure what I was after, the main benefit being that when, for instance, I configured ether4 and ether5 to belong to vrf2, then dynamically added routes involved vrf2 table (and not main table anymore).
May I ask about the purpose? I am curious if any of the endpoints would ever know whether their packets went through that 4 step NAT process instead of a 1 step NAT process?
I’m not sure I correctly understood your question (I’m sorry for that) but my purpose is to simulate in a lab with a single box, up to four different mostly independent NAT routers.