Lorentz proposed the equations for the Lorentz transformation (aka Lorentz contraction) before Einstein figured out how to derive them from the assumption of fixed c. These formulas were derived by requiring that Maxwell's equations be invariant under a change of coordinates.
The non-SR interpretation is that there's an absolute, fixed ether, but when you move relative to the ether you get a contraction effect which results in your length and time measurements getting messed up just right. Thus, the Michelson-Morley experiment got a null result because the length of each arm of the apparatus changed as it was rotated relative to its motion through the ether.
Now the transformation equations are the same ones that Einstein derived from the assumption that c is fixed in all frames. So the actual predictions are the same: the theories are effectively equivalent.
So, why prefer Einstein's interpretation to Lorentz's? There are, I believe, three reasons.
First, insisting that Maxwell's equations be invariant is a good idea, but introducing the Lorentz transformation on those grounds is a bit ad hoc. Insisting that there is an ether and/or an absolute frame of reference, but it isn't detectable because objects compress when they move relative to it is rather complicated and "just so" vs. simply assuming c is constant. So out comes Occam's Razor.
Second, I think Einstein's SR is conceptually more productive. It leads more naturally to GR than a theory based on Lorentz alone would have. Now, people who insist on the ether like GR even less than SR, so they probably wouldn't have a problem with this, and in comes the VSL theory instead of GR. But just classical electrodynamics by itself is simpler in the formalism of special relativity.
Third, and I think this is very important, we need to think about what the definition of time and distance intervals are. Intervals in time and space must be defined operationally - that is, distance is defined by mechanical comparison to some length standard or device like a ruler, or a certain number of wavelengths of a certain sort of light, or something like that. Time intervals are measured by comparisons to mechanical clocks. Einstein's interpretation takes more seriously the operational nature of measurement - in fact, Einstein's derivation of the Lorentz transformations are based on the operational nature of time and distance measurements. Defining an absolute frame and a distortion thereof treats distance and time intervals as things that have an abstract existance independent of the measuring process and then "fixes" measurement problems with a hack. It's not operational.
Now, I don't know so much about the relationship between GR and VSL. My understanding of VSL is that its based on the idea that warps in spacetime can be treated mathematically like a variation in index of refraction (the optical analogy) and VSL is based on that. This seems like a neat idea. Again, I believe the mathematics ends up being equivalent, so there's no falsifiable way to decide between them - they're just two different interpretations of the same theory and, operationally, all the wierdness that happens in one theory will be present in the other (like differences in clocks, etc).