The increasing number of sensor-equipped mobile devices enables new applications for communication within crowds of people. Ranging from monitoring services that provide insights into the crowd's behavior to fully fledged messaging applications for end users, all such applications require communication protocols that are tailored to the characteristics of a mobile ad hoc network (MANET). A common communication scheme for such dynamic networks is the publish/subscribe (pub/sub) paradigm, that offers temporal and spatial decoupling. However, pub/sub protocols for MANETs are designed and evaluated under very restricted conditions regarding the expected mobility, number of users, application workload, and application requirements, targeting rather specific scenarios. In reality, such conditions are subject to change, especially considering people's behavior during crowded events. In this work, we propose a framework that enables dynamic transitions between different pub/sub protocols based on the currently monitored conditions or caused by an external trigger. We analyze the behavior during transitions and the overhead introduced by our framework through extensive simulations. By using simple pub/sub protocols and switching between them based on the observed conditions, we are able to maintain good service quality at reasonable overhead, thereby enabling applications to operate in dynamic conditions representative of mobile crowds.