This research gauges the ability of deep reinforcement learning (DRL) techniques to assist the optimization and control of fluid mechanical systems. It combines a novel, "degenerate" version of the proximal policy optimization (PPO) algorithm, that trains a neural network in optimizing the system only once per learning episode, and an inhouse stabilized finite elements environment implementing the variational multiscale (VMS) method, that computes the numerical reward fed to the neural network. Three prototypical examples of separated flows in two dimensions are used as testbed for developing the methodology, each of which adds a layer of complexity due either to the unsteadiness of the flow solutions, or the sharpness of the objective function, or the dimension of the control parameter space. Relevance is carefully assessed by comparing systematically to reference data obtained by canonical direct and adjoint methods. Beyond adding value to the shallow literature on this subject, these findings establish the potential of single-step PPO for reliable black-box optimization of computational fluid dynamics (CFD) systems, which paves the way for future progress in optimal flow control using this new class of methods.