In massive MIMO, the CSI acquisition at the BS is the most challenging issue. Without accurate CSI, NOMA can not perform well. NOMA has an advantage in CSI acquisition. It allows gradual CSI refinement through iterative processing.
The NOMA is as a multiple access technique relying on the amplitude division multiplexing. It is used to increase the spectral efficiency of the wireless communication system. Mostly one can use use two signal levels, so, it is probably not effective for increasing the spectral efficiency.
Massive MIMO is very effective space division multiplexing system with high multiplicity factor the available bandwidth. So, it is an effective space division multiplexing it is much more effective than NOMA.
Both techniques can be combined but the system may get more complicated which render the combination less feasible. However, one can try to study such combined system ans see its yield, specially both techniques rely on the knowledge of channel sate for signal detection.
NOMA is useful when two (or more) users have similar channel directivity, so that they can be covered by the same beam, but different pathlosses. One of the key properties of massive MIMO is called "favorable propagation" and means that the chance of having two users with similar channel directivity reduces as you add more antennas. If you go up in frequency (and keep the aperture fixed), then the beams will be even more directive.
In other words, more antennas and higher frequency are two properties that reduce the NOMA gain until the point where it disappears. At terahertz frequencies, I doubt that NOMA will be useful.
This is not a bad thing, it is a good thing! It means that using massive MIMO is sufficient - no need for an additional NOMA layer.
I think as the number of users increases the probability to find two users aligned with each other increases one is near and the other is far. So, they will be served by one beam. So, whenever the the number of users increases the spectral efficiency using NOMA increases.
As the number of users in coverage area increases, the odds that's two or more users have similar channel directivity increases as well, which can be covered with the same beam, which eventually increases the spectral efficiency of NOMA system. Additionally, Tera-hertz communication link fits well with this scenario, due to its high carrier frequency, which require direct links with UEs to establish the link.