Reliable Server Pooling (RSerPool) is the IETF's new standard for a common server redundancy and session failover framework to support availability-critical applications. Server pools are maintained by redundant management components denoted as registrars. These registrars monitor the availability of servers in the pool and remove them in case of failure. Furthermore, they synchronize their view of the pool with other registrars to provide information redundancy. In this paper, we first illustrate the implications of registrar redundancy on the performance of RSerPool systems. After that, we present an optimization approach for the server pool management, which improves the management performance in case of registrar problems like hardware failures or Denial of Service attacks. The performance of our approach is evaluated in real life using PlanetLab measurements.