Recent crystallographic data unambiguously demonstrate that neither Ar'GeGeAr' nor Ar'CrCrAr' molecules adopt the expected linear (VSEPR-like) geometries. Does the adoption of trans-bent geometries indicate that Ar'MMAr' molecules are not "maximally bonded" (i.e., bond order of three for M = Ge and five for M = Cr)? We employ theoretical hybrid density functional (B3LYP/6-311++G**) computations and natural bond orbital-based analysis to quantify molecular bond orders and to elucidate the electronic origin of such unintuitive structures. Resonance structures based on quintuple M-M bonding dominate for the transition metal compounds, especially for molybdenum and tungsten. For the main group, M-M bonding consists of three shared electron pairs, except for M = Pb. For both d- and p-block compounds, the M-M bond orders are reflected in torsional barriers, bond-antibond splittings, and heats of hydrogenation in a qualitatively intuitive way. Trans-bent structures arise primarily from hybridization tendencies that yield the strongest sigma-bonds. For transition metals, the strong tendency toward sd-hybridization in making covalent bonds naturally results in bent ligand arrangements about the metal. In the p-block, hybridization tendencies favor high p-character, with increasing avidity as one moves down the Group 14 column, and nonlinear structures result. In both the p-block and the d- block, bonding schemes have easily identifiable Lewis-like character but adopt somewhat unconventional orbital interactions. For more common metal-metal multiply bonded compounds such as [Re2Cl8](2-), the core Lewis-like fragment [Re2Cl4](2+) is modified by four hypervalent three-center/four-electron additions.