The unusual autodetachment spectrum of metastable ${\mathrm{He}}_{2}^{\mathrm{\ensuremath{-}}}$ is studied theoretically, using high-level coupled-cluster methods and hybrid numerical and Slater orbital basis sets. By obtaining accurate curves for the repulsive wall of the ground ${\mathrm{He}}_{2}$(X $^{1}\mathrm{\ensuremath{\Sigma}}_{\mathrm{g}}^{+}$) state, the excited ${\mathrm{He}}_{2}$(a $^{3}\mathrm{\ensuremath{\Sigma}}_{\mathrm{u}}^{+}$) state, and the metastable ${\mathrm{He}}_{2}^{\mathrm{\ensuremath{-}}}$${(}^{4}$${\mathrm{\ensuremath{\Pi}}}_{\mathrm{g}}$) state, we are able to provide an alternative explanation for the experimental observations, which had cast doubt on the veracity of the accepted curve for the repulsive part of the ${\mathrm{He}}_{2}$(${\mathrm{X}}^{1}$${\mathrm{\ensuremath{\Sigma}}}_{\mathrm{g}}^{+}$) potential. We attribute the experimental peak at 15.78\ifmmode\pm\else\textpm\fi{}0.13 eV to transitions of vibrationally excited states (\ensuremath{\nu}=2 and higher) of ${\mathrm{He}}_{2}^{\mathrm{\ensuremath{-}}}$ to the ${\mathrm{He}}_{2}$(X $^{1}\mathrm{\ensuremath{\Sigma}}_{\mathrm{g}}^{+}$) continuum, since the v=0 transition would have a value 1 eV less. An error of this size is considered to be far outside the error bars for highly accurate correlated ab initio calculations. The electron affinity of the $^{4}\mathrm{\ensuremath{\Pi}}_{\mathrm{g}}$ state of the anion (measured relative to a $^{3}\ensuremath{\Sigma}_{u}^{+}$) is computed to be 0.201 and 0.212 eV for different basis sets, compared to an experimental value of 0.175\ifmmode\pm\else\textpm\fi{}0.032 eV.