This work presents a novel control strategy for systems with actuation delays with known stochastic distribution, which improves upon previously proposed deadline-driven and event-driven strategies. In the event-driven strategy, the control input is immediately updated after the delay, whereas in the deadline-driven strategy, the actuation is updated in a periodic fashion, where the sampling period sets a deadline; if the delay is larger than this deadline, the actuation is not updated. Our method switches between these two strategies and guarantees better performance, in a linear-quadratic-Gaussian sense, than either method considered separately. An extension of the novel method with a deadline-optimization scheme is shown to improve the performance even further. Simulation results illustrate the effectiveness of the proposed methods.