It has been a puzzle for over two decades that the enhancement of oxygen diffusion in α-Al_{2}O_{3} ,with respect to the amount of Mg doping, is several orders of magnitude less than expected. The standard model, which envisages that transport is mediated by oxygen vacancies induced to compensate the charge of Mg 2+ ions substituting Al 3+ ions, has not been able to explain this anomaly. Here, we report a detailed study of populations of point defects and defect clusters in Mg-doped α-Al_{2}O_{3}. By taking into account calculated defect formation energies from the literature, the condition of charge neutrality, and the environmental parameters (chemical potentials) under which the anomalous trend in oxygen diffusivities were previously observed, we are able to arrive at an explanation. A non-linear relationship between Mg concentration in the system and key native point defects, which serve as mediators of self-diffusion in α-Al_{2}O_{3_ , is predicted: the concentrations of such defects increase much more slowly in the supersaturation regime than in the pre-saturation regime, matching the anomalous result previously observed in α-Al_{2}O_{3} . We identify the reason for this as buffering by positively charged Mg interstitials and Mg–oxygen vacancy clusters, which compensate the negative charges of Mg substitutional defects (Mg^{1−}Al ). This study answers part of the long-standing question about self-diffusion in alumina, referred to by Heuer and Lagerlöf in 1999 as the Corundum Conundrum.