This work explores dynamical couplings between global enzymatic fluctuation patterns and the catalytic fitness of a model enzyme system. A genetic algorithm drives the evolution of a population of enzyme molecules toward greater catalytic fitness by modifying via recombination and mutation events the interaction strengths between enzymatic subunits. We've identified, through a combination of molecular dynamics simulation and an approximate normal-mode analysis, two features that are tuned for increased catalytic efficacy throughout the evolutionary process. First, the average distance between substrate and the enzymatic subunit responsible for the chemistry is optimized. Second, the fraction of rate-promoting oscillations at the reaction coordinate is maximized. In other words, enzyme evolution in our model system proceeds by tuning geometric constraints and effectively eliminating catalytically nonproductive fluctuations. This modulation at the reaction coordinate results from global fluctuation patterns that are established by the spatial mix of loose and stiff enzymatic domains. The optimized geometry and appearance of rate-promoting oscillations at the reaction coordinate are emergent properties of the fluctuating enzyme system: they are not built into the model or selected for by the genetic algorithm. As is the case for actual enzyme molecules, catalytic fitness in our model system is sensitive to single-point mutations that affect global collective motions.