The techniques and mechanisms described herein are directed to controlling communication paths between conference members in a decentralized multi-party conference that supports a full mesh architecture. Several check points are tested in order to handle concurrency issues and to ensure quality of service. First, a check is made to determine whether any pending communication paths exist between the two systems. Second, a check is made to determine whether one of the systems is in the process of leaving the multi-party conference. Third, a check is made to determine whether a maximum number of conference members already exist in the conference. Upon failing any of these checks, the communication path between the two systems fails. Upon success, the new system initiates the joining process with each of the other members in the conference in order to maintain the full mesh architecture.