Packet monitoring arguably needs the flexibility of open architectures and active networking. A significant challenge in the design of open packet monitoring systems is how to effectively strike a balance between flexibility, safety and performance. In this paper we investigate the performance of FLAME, a system that emphasizes flexibility by allowing applications to execute arbitrary code for each packet received. Our system attempts to achieve high performance without sacrificing safety by combining the use of a type-safe language, lightweight run-time checks, and fine-grained policy restrictions. Experiments with our prototype implementation demonstrate the ability of our system to support representative application workloads on Bgit/s links. Such performance indicates the overall efficiency of our approach; more narrowly targeted experiments demonstrate that the overhead required to provide safety is acceptable.