Well, there is a point in any LOD system where there's a break even moment between faster drawing time and longer calculating, read CPU performance, time. The goal obviously is to find the right balance between the two and, not to forget, to find the right distance from the viewer where the change, or changes, depending on the number of LOD phases, takes place.
I can imagine that last one can be tricky in StarMade because of the sizes of some builds. It'll be interesting to see how that works out when implemented.
Greets,
Jan