A two‐stage multi‐server tandem queue with two types of processed customers is analyzed. The input is described by the Marked Markovian Arrival Process (MMAP). The first stage has an infinite number of servers while the second stage has a finite number of servers. The service time at the both stages has an exponential distribution. Priority customers are always admitted to the system. Non‐priority customers are admitted to the system only if the number of busy servers at the second stage does not exceed some pre‐assigned threshold. Queueing system’s behavior is described in terms of the multi‐dimensional asymptotically quasi‐Toeplitz continuous time Markov chain. It allows to exploit a numerically stable algorithm for calculation of the stationary distribution of the queueing system. The loss probability at the both stages of the tandem is computed. An economic criterion of the system operation is optimized with respect to the threshold. The effect of control on the main performance measures of the system is numerically demonstrated.