Recent measurements indicate that the rate coefficient for Na+O2+M→NaO2+M is three orders of magnitude faster than previously reported. Use of this new rate coefficient in current models will lead to Na concentrations in the upper mesosphere significantly lower than the observations. We propose that the restoration of the proper Na concentrations necessitates the reduction of NaO2 by O, particularly at night when photolysis of NaOH or NaO2 become ineffective Na sources.