The accuracy of the variable reaction coordinate (VRC) implementation of transition state theory (TST) is investigated for the bimolecular reaction of H with O-2 via direct comparisons with quantum scattering theory for J=0, classical trajectory simulations for a wide range of J, and experimental canonical rate constants. The DMBE TV potential energy surface of Varandas and co-workers is employed in each of the theoretical calculations. The first two comparisons indicate that the VRC-TST approach overestimates the cumulative reaction probability (CRP) for this reaction by a factor of 2.3, roughly independent of E and J for moderate energies. The trajectory simulations further indicate that this failure of TST is primarily the result of the rapid redissociation of a large fraction of the initially formed HO2. An estimate for the quantum CRP on the basis of the combined dynamical and statistical results is seen to provide a useful alternative to the more standard quasiclassical trajectory estimates. A thermal averaging over the E and J-dependence of the TST estimates for the CRP provides canonical rate constants, k(T), which, when corrected for the above-mentioned overestimate, are still a factor of 1.7-2.0 times greater than the experimental data. This discrepancy is most likely the result of either (i) inaccuracies in the DMBE IV surface and/or (ii) an overestimate of the contribution to the reactive flux from the nearly degenerate first excited state in the exit channel region.