I've read in multiple locations that GR is necessary to properly explain a co-rotating perspective/observer of the experiment. Additionally, SR can explain the experience of non-inertial observer. The above two statements seem somewhat conflicting and leads to my question. Why is it that GR is required to properly predict the Sagnac effect. What is it that causes the SR explanation to be approximate but inadequate for an exact answer? I'm not looking for the calculation, I'm looking for the reasoning required to say: GR takes X into account while SR does not.