The thermal neutron elastic-scattering cross-section data for 16O used in various modern evaluated nuclear data libraries were reviewed and found to be generally too high compared with the best available experimental measurements. Some of the proposed revisions to the ENDF/B-VII.0 16O data library and recent results from the TENDL system increase this discrepancy further. The reactivity impact of revising the 16O data downward to be consistent with the best measurements was tested using the JENDL-3.3 16O cross-section values and was found to be very small in MCNP5 simulations of the UO2 and reactor-recycle MOX fuel cases of the American Nuclear Society Doppler defect numerical benchmark. However, large reactivity differences of up to [approximately]14 mk (1400 pcm) were observed using 16O data files from several evaluated nuclear data libraries in MCNP5 simulations of the Los Alamos National Laboratory highly enriched uranium (HEU) heavy water solution thermal critical experiments, which were performed in the 1950s. The latter result suggests that new measurements using HEU in a heavy water-moderated critical facility, such as the ZED-2 zero-power reactor at the Chalk River Laboratories, might help to resolve the discrepancy between the 16O thermal elastic-scattering cross-section values and thereby reduce or better define its uncertainty, although additional assessment work would be needed to confirm this.