The benefits of using a mobile base station to prolong sensor network lifetime have been well recognized. However, due to the complexity of the problem (time-dependent network topology and traffic routing), theoretical performance limits and provably optimal algorithms remain difficult to develop. This paper fills this important gap by contributing some theoretical results regarding the optimal movement of a mobile base station. Our main result hinges upon two key intermediate results. In the first result, we show that a time-dependent joint base station movement and flow routing problem can be transformed into a location-dependent problem. In the second result, we show that, for $(1- varepsilon)$ optimality, the infinite possible locations for base station movement can be reduced to a finite set of locations via several constructive steps [i.e., discretization of energy cost through a geometric sequence, division of a disk into a finite number of subareas, and representation of each subarea with a fictitious cost point (FCP)]. Subsequently, for each FCP, we can obtain the optimal sojourn time for the base station (as well as the corresponding location-dependent flow routing) via a simple linear program. We prove that the proposed solution can guarantee the achieved network lifetime is at least $(1- varepsilon)$ of the maximum (unknown) network lifetime, where $varepsilon$ can be made arbitrarily small depending on the required precision.