One of the problems faced by BGP routers today is the BGP churn, i.e. the number of messages that they have to process. Measurement studies have shown that a small fraction of the prefixes are responsible for a large fraction of the BGP messages that are received by BGP routers. Furthermore other studies have shown that the destination prefixes that carry lots of traffic usually have stable BGP routes.
To protect the CPU of the BGP routers, network operators have pushed vendors to implement techniques that allow to reduce the number of BGP updates that need to be processed by a router. One of these techniques is Route Flap Damping. This algorithm associates a penalty to each prefix based on the number of BGP messages exchanged in the past for this prefix and prefixes with a too large prefix are removed from the routing table for some time. Route Flap Damping was initially recommended and RIPE proposed a set of parameters. However, a more detailed analysis of Route Flap Damping showed that it interacted badly with the path explorations that could legitimately occur due to the path vector nature of BGP. For this reason, network operators changed their recommendations on the usage of Route Flap Damping. The current set of recommendations may be found in a recent Internet draft : draft-shishio-grow-isp-rfd-implement-survey-00

A recent paper, Route Flap Damping Made Usable, proposes a new set of parameters and suggests to re-enable Route Flap Damping on BGP routers.
Given the amount of updates that badly configured ASes continue to generate, I'm not personally convinced that using Route Flap Damping is still the best approach. I'd suggest to block the prefixes that flap a lot for a much longer time (e.g. an entire day or more). This would be a strong message to network operators to encourage or perhaps force them to take measures to reduce the amount of churn on their own prefixes.