Dugald is pretty much on the ball with his recent contribution. I
don't think we do see regional transitions from amphibolite- to
granulite-facies (with breakdown of biotite and hornblende) in which
the equilibria responsible are subsolidus things. fluid-absent
melting seems to be the rule, and seems certain to be the main
mechanism by which granitic magmas are generated.
Clemens J.D., Watkins J.M. (2001) The fluid regime of
high-temperature metamorphism during granitoid magma genesis. Contrib
Mineral Petrol 140: 600-606
Those curious about the effects different fluid regimes (and
compositions) on partial melting, degrees of segregation (physical or
chemical) between melt and restite and the affects of different
retrograde paths might care to have a look at the models presented in:
Clemens J.D., Droop G.T.R. (1998) Fluids, P-T paths and the fates of
anatectic melts in the Earth's crust. Lithos 44: 21-36
This treatment is pretty comprehensive. OK it's a bit full of
theoretical phase diagrams (P-T, T-X and P-X) to be an easy read, but
I still reckon there's something practical to be gleaned from this
kind of stuff. Summary sections deal with what we should actually see
in the rocks.
That's enough shameless self advertising,
PS I have no reprints left of either of these two papers.