This paper provides a control method for autonomously tracking moving targets with slow and delayed visual feedback. Such capabilities underpin robotic applications ranging from transportation to medical systems. Compared to classic position-based visual servo, the proposed approach achieves 95% reduction of tracking errors when partial knowledge of the target dynamics is known. The result is achieved via a tripartite approach including (1) a velocity control of the robot that ensures exponential reduction of the visual feature error even subject to a maneuvering target, (2) a pose and velocity estimation that considers the dynamics of the moving target, and (3) a multirate information recovery that recovers fast motion estimates from slow and delayed visual feedback.