Hypothetically, one could fluorinate to 12CF4/13CF4/14CF4 to circumvent 18O problem. We measure 36SF6 (200 ppm abundance) at 0.2 permil precision. Tens of ppm 14CF4 might give good enough precision depending upon your goal.
You could probably use laser spectroscopy; we measure 13CH3D that is <6 ppm fractional abundance at 0.3 permil precision (Anal Chem, 86, 6487). CO2 has a fairly strong IR absorption for asymmetric stretch.
13C-magnetic isotope effect might produce large anomalous relationships that might be captured. Interesting to think about.