I think another way to say it is that you would need to alter the speed of light if you wanted to get a curved path in a flat spacetime. If instead you curve the spacetime, then you can have a constant speed of light. So if this is correct, then curvature is seen as a way to achieve SR results locally within the context of GR. Alternatively, you may allow the speed of light to vary, and not curve the spacetime, but that's not the usual formulation. I may be wrong here, I'm trying to understand the possible pedagogies. Also, in terms of the vacuum, I would certainly expect that two observers in relative motion to each other would nevertheless observe the same vacuum, so how zero-point oscillations relate to a local ether is not obvious.