@atyy and PAllen, the distinction between Galilean space-time and Minkowski space-time is the latter presumes a space-like foliation with separate Euclidean temporal and spatial metrics whereas the latter contains no such foliation and presumes the Minkowski metric. From here one can demand that equations (or "the laws of physics") be Lorentz (or Poincare) covariant and get SR. Why are inertial frames required for any of this? Equations will be Lorentz covariant as long as they are written in terms of spinor or tensor representations of the Lorentz group which is a completely frame-independent condition. After the dust settles we can simply define an inertial frame in terms of zero rotation and acceleration. I honestly don't see any need to talk about inertial frames before the dust settles.