How closely should I match the index of 'index matching fluid' to that of a ground glass fused silica cube such that a laser incident on the surface does not scatter?
It's gonna scatter anyway. You may try to match the indices as closely as possible and then try to vary the temperature of the cube-oil-(some flat glass) assembly, hoping to match the temperature-dependent indices. But I'd guess you'll never get rid off of some nasty interference effects in the far field. Depending on your application, it might help to pass the beam through some filtering pinhole or do other tricks...