GPU rendering might be the future, but in my opinion it is in its infancy and that is an understatement. First of all a rendering package is sold as a product and the main buyers are probably commercial entities rather than hobbyists. As such the rendering package needs to address the main user base, rather than looking at the smaller percentiles of the distribution of the user base.
You are wrong, I have given up on hair rendering in Lightwave, it is outmatch by the GPU rendering with blenders hair, seems it´s like 20 times faster.. though even the cpu hair rendering is faster than lightwave native CPU.
And you are wrong with volumetric gi bouncing, it´s absent even in native vdb volumes or volumetric items, activating gi for volumetrics with cpu is very very slow compared to blenders cycles voluemtrics with GI and a proper functionial volume bouncing, I have demonstrated that with the Disney VDB asset and other stuff, so I dare to state you are not aware of what you are talking about here.
Besides, look at those using Lightwave and not blender to do volumetrics or instancing etc, like Lightwave guru, how much isn´t he working with octane, and what is octane if not a GPU renderer, it´s the solution most seem to have to buy in to to overcome the slow CPU processing for some tasks...that or invest in very very expensive cpu´s