How about assembling a GLSL team?
I know I'm only on the outside looking in, but it appears that of all the features, GLSL is either the hardest to implement, or takes the most time for piglit/etc verifications, the most in-depth; I don't know what would be the proper way to describe how it looks from out here.
But It seems like GLSL 3.2 and 3.3 have been "in progress" fffffoooooorrrrreeeevvvveeeerrr, so why not keep whomever is the core development team for MESA GLSL on GLSL - when they get done with 3.2 and 3.3, in light of the Fred Brooks problem, put them right on GLSL 4.0.
When the GLSL team is done with GLSL 4.0, put them on GLSL 4.1.(Skipping the rest of the OpenGL features)
When done, put them on GLSL 4.2.
When done, put them on GLSL 4.3.
Then GLSL 4.4.
Then GLSL 4.5. (Yes, I know it doesn't exist yet. But it probably will by the time all these other GLSLs are done)
Just a random thought from a complete outsider looking in. A dedicated GLSL Mesa team seems to make a lot of sense. The rest of the features will catch up.