As Gray says, returns obviously diminish in empirical terms.
But there's another way of looking at this - I know you can't put a percentage on how a system sounds, but humour me as a way of illustrating my point. Let's say someone has a modest system, and we gauge it as sounding 60% of real (yes, I know this opens up more questions than it might answer...). Suppose a system based upon double the budget gets to 80%, that's obviously a diminishing return. But a doubling of budget has also halved the gap between what the system can deliver and 'reality'. In that respect, my rather contrived example does deliver. A doubling of budget halves the deficit to reality.
I think that sometimes relatively small improvements can be the things that maximise listening pleasure. It depends upon how demanding/picky (delete as applicable) the listener is.