This paper addresses real-time optimization strategies which can be readily implemented in industrial polymerization processes, even in case they show very fast dynamics. At the upper layer dynamic and steady-state real-time optimizations (D-RTO and RTO) are suggested and compared. A novel multistage formulation for the real-time dynamic optimization problem is introduced. It relies on a purely economic objective without additional stabilizing terms and facilitates an integrated treatment of a sequence of alternating dynamic and stationary operational stages. A case study shows that closed-loop D-RTO allows reducing off-spec material as well as exploiting or rejecting disturbances to maximize overall profit. The computational delay not only determines closed-loop performance but also significantly impacts the profit margin. The results indicate that even the simplest variant of the investigated strategies can significantly improve economic performance, since the transitions can be completed much faster than in current industrial practice.