Weight and volume are surrogates for the size of the connectome, just not perfect ones (synaptic pruning in adolescence, tumors etcetera).

Healthy versus diseased as a binary choice is too simplistic. While the example I used was indeed of a rather severe disease, it also applies to e.g. elderly brains versus mid-life brains, or kids versus adults.

On any major axis I can think of (age, "diseases"), observing a smaller brain area (also: a loss of neurons) if anything predicts for smaller other brain areas (also: loss of neurons elsewhere). A positive Pearson's r, not a negative one.

Why would other areas shrink? Because cranial capacity is limited? You'd think so, but I've never come across anything of the kind (other than the niche cases mentioned). If you remember this as a typical phenomenon, I'd appreciate the source.