If MS does anything about this issue at all it will be because of Naples if they even need this done for that unit.
It's good enough on desktop, but likely not good enough on server is my guess. If we get anything out of a change it will be because of that trickle down. But I doubt it will add up to much with only two clusters on Ryzen.
The only way MS can do something about this is to make a NUMA like system for AMD's CPUs, I highly doubt they will do something like this because it will increase complexity of programming for it to an extreme level.
NUMA programming is not easy because workloads have to be sufficient enough to saturate each NUMA node, games in particular are not good with this, as distributing the main threads is quite complex with all the interdependence requirements.