The reliability of density functional theory (DFT) for the prediction
of equilibrium molecular properties is assessed for several challenging
molecules (F_{2}^{+}, O_{2}^{+},
O_{2}, O_{4}^{+}, F_{4}^{+},
and NO_{3}) which tend to artifactually break spatial symmetry
in the Hartree-Fock approximation. DFT appears to be among the more
robust methods for such cases, but it is not immune to anomalous
property predictions. The BLYP and BPW91 functionals fail
dramatically for rectangular O_{4}^{+} and
F_{4}^{+}, where they are unable to find a bound
minimum.