According to this author, http://www.math.uic.edu/undergraduate/mathclub/talks/Weeks_AMM2001.pdf, a locally Minkowski spacetime with a nontrivial global topology may have a preferred inertial frame, in the sense that hypersurfaces of constant time can only be defined using particular time coordinates. Is he right? Is there any reason such spaces should be considered unrealistic? This seems to cast doubt on the basic (at least the philosophical) premise of relativity... Also, does this point have relevance within GR, in particular for our universe? Does it say something important about the CBMR frame, as the writer claims?