Even if it's parallelizable, you have all sorts of single core embedded CPUs, in which case it isn't true either. So I get your point.
But something like making a whole major release whose primary goal is to reduce the memory consumption of Gnome to 10% is prone to leaving Gnome in the dust bin of desktop history - you only have so much resources, and when you try this kind of optimization it becomes all you do, you can't really add features.
A noble goal, hardly practical. Notice the evolution of all successful software, from Sendmail to Windows NT. MOAR FEATURES!