We investigate the effect of packet transmission, i.e., noncontinuous transmission of light, on polarization control for various traffic patterns and loads. Time between feedback signal samples, where a packet is present, i.e., when feedback is available, becomes a statistical quantity in asynchronous networks. By simulations we characterize the polarization state fluctuations occurring between valid samples when packet interarrival time is random and exponentially distributed. We quantify dynamic polarization control performance in this situation. Worst-case polarization state fluctuations between valid samples can reach 30?? and beyond, compared to max ~2.5?? rotation in a sample time interval with continuous light. As a rule of thumb, polarization control is possible when packet lengths are smaller than or comparable to the time scale of the polarization fluctuations, even for traffic loads as low as 10%. Examples are links carrying asynchronous IP packet traffic at bit rates of 2.5 Gb/s or higher, with worst-case polarization fluctuations on the order of microseconds.