Blender's dogfooding efforts were a powerful force in both proving that the software was good enough (to make movies) and demonstrating how it could be better (for the animators who made the movies). I'm not aware of any such efforts for GIMP. What I have seen is roadmaps of what improvements they know they need, and timelines of years to improve core infrastructure to the point that their actual goals are within reach. Meanwhile, other software projects (both FLOSS and commercial) race past them and move on to more exotic and exciting features.