Mobile Ad Hoc Networks play a significant role in upcoming network architectures like the Internet of Things. Due to their disorganized nature and high mobility of nodes, these networks are prone to partitioning and volatile network topologies. Both effects create unique requirements for the deployment of distributed services. Since the service provider cannot control a node's physical properties, we focus on finding solutions at the application layer. In a distributed service such as distributed caching, it is necessary that a node dynamically adapts its storage to the changing network situation to maintain the overall redundancy in the system, even in the case of partitioning. Our work proposes a failure model for network partitioning and simultaneously occurring node movements. We use this model to evaluate several improvements to state-of-the-art redundancy management schemes to make the distributed caching network resilient against network partitioning. We can reduce the storage costs by up to 66% while at the same time minimizing data transmissions. In addition, we investigate how the movement of nodes influences the requirements and show that in high mobility scenarios, the storage cost is up to 26% lower than in a static scenario.