ER: send queue grows

Hi all. We have 3 servers using ER: - Server A, primary, IDS 9.21.UC4,Linux - Server B1, target, IDS 9.21.UC4, Linux - Server B2, target, IDS 9.30.FC3, Solaris 8 The 3 replication servers have been created with the send and receive queues in a dbspace called "physdbs". On each server, this dbspace is only used by the replication queues. Every thing seems to be OK (transactions are applied on both target servers), but we observe that the "physdbs" dbspaces continuously fills on server A and on server B1. The dbspace does not fill on server B2. Could you tell me what happens to servers A and B1 ? A known problem with ER 9.21.UC4 ? Do I Need to purge the queues myself on these servers ? The only way I have found to empty the physdbs is to "cdr delete server"... Note that we have defined two separate sets of replicates: one for A-B1 and one for A-B2. Thank you for your help, Philippe