Technote 1093915 outlines some information which is nice to remember when using @Formula to access databases in clusters. After considering this technote you might want a scheduled LotusScript or Java agent to discover the replica id of the target database and store the replica id in a profile document for usage by formula agents. This approach is usable if you want failover to work correctly and doesn’t want to hardcode the replica id into agents which I always consider bad practice.