Using matrices for points wields a lot of power, but at the same time you need four times the memory. That is why usually simple single vector positions are used for anything with a large point count. Even for splines you mostly try to get away with less than a matrix, i.e. broken tangents only increase memory consumption by a factor of two, not three like a matrix.
Neutron doesn't change anything here, some things just become a bit more visible than in the past.
What Neutron does differently is that you can already attach specific attributes to vertices, where in classic Cinema 4D you have to add tags this kind of information is now a direct part of the geometry information. At this time these attributes are a bit limited, you can only use types that are offered within the Geometry Property Set node. On the other hand, where possible, all modeling operations interpolate/copy those attributes if the topology changes.