The Delay/Fault-Tolerant Mobile Sensor Network (DFTMSN) has been proposed recently for pervasive information gathering. DFT-MSN distinguishes itself from conventional sensor networks by several unique characteristics such as sensor mobility, loose connectivity, and delay/fault tolerability. The mainstream approaches for sensor networking/ communication cannot be applied in DFT-MSN directly. In this paper we propose a Replication-based Efficient Data Delivery (RED) scheme based on erasure coding technology tailored for DFT-MSN. RED consists of two key components for data transmission and message management, respectively. The former makes decision on when and where to transmit data messages according to the delivery probability. The latter decides the optimal erasure coding parameters (including the number of data blocks and the needed redundancy) based on its current delivery probability, in order to achieve the desired data delivery ratio while minimizing overhead at the same time. Extensive simulation has been carried out for performance evaluation. Compared with other approaches, the proposed RED data delivery scheme achieves high message delivery ratio with low transmission overhead and data management complexity.