According to one influential paradigm, malignant phenotypes characterizing the "hallmarks of cancer" arise in part via natural selection acting on genetically diverse clones within a tumor. Among these hallmarks, the angiogenic switch is one of the most difficult to explain using an evolutionary narrative. While neoangiogenesis clearly benefits tumor cells, the signal creating it is a public good and therefore susceptible to free-riders. Previous modeling studies predicted that these free-riders can invade, damage and perhaps destroy developing tumors, growing as a tumor-on-a-tumor, or hypertumor. The open question becomes, why are hypertumors apparently rare? Here we show, using more realistic extensions of the original models, that selection favoring free-riding is expected to be overwhelmed by genetic drift in most cases. Adaptive dynamics analysis of a deterministic model of the energetic costs and benefits of angiogenesis and proliferation predicts the existence of an evolutionary stable (ESS) angiogenesis commitment, but this ESS is always a repeller. The expectation, then, is runaway selection for extreme vascular hypo- or hyperplasia. However, the selection gradient is very shallow compared to that for other traits, specifically proliferation. Therefore, evolutionarily unfavorable angiogenesis phenotypes may still invade if they are coupled to even marginally more favorable proliferation strategies through a mechanism logically identical to linkage disequilibrium. A simulation of this evolutionary theater predicts that this disequilibrium mechanism dominates the evolution of the angiogenic switch. We predict, then, that angiogenesis arises as an evolutionary rider on the back of selection for proliferative potential and other malignant hallmarks.