TCP throughput instability is a well-known phenomenon in IEEE 802.11 multi-hop ad-hoc networks. However, we find that this problem is not restricted to TCP traffic only, but also occurs in UDP traffic. The associated throughput oscillations are not acceptable for real-time applications such as video conferencing and voice over IP. This paper re-defines this throughput fluctuation as a "re-routing instability problem" since it is caused by the triggering of the re-routing function. In particular, we show that the throughput instability is mainly induced by re-routing, not the binary exponential back-off of the IEEE 802.11 MAC protocol. Turning off the re-routing function, for example, eliminates the problem. We believe that this is the first paper in the literature to study this phenomenon in the context of rerouting instability. We propose to modify the ad-hoc routing protocols with a "don't-break-before-you-canmake" strategy. The scheme does not require modif...