Reducing the Cost of Cycle-Time Tuning for Real-World Policy Optimization