Overload in a packet-based network can be prevented by admitting or blocking new flows depending on its load conditions. However, overload can occur in spite of admission control due to unforseen events, e.g., when admitted traffic is rerouted in the network after a failure. To restore quality of service for the majority of admitted flows in such cases, flow termination has been proposed as a novel control function. We present several flow termination algorithms that measure so-called pre-congestion notification (PCN) feedback. We analyze their advantages and shortcomings in particular under challenging conditions. The results improve the understanding of PCN technology which is currently being standardized by the Internet Engineering Task Force (IETF).