Cooperative Multi-Point (CoMP) has emerged as a new paradigm to improve both average cell and cell edge throughput in cellular networks. However, the performance is significantly degraded due to Out-of-Group Interference (OGI). One way to mitigate OGI is to restrict the interfering signal to lie inside the null space of the unintended receiver. Yet, accurately tracking this null space is a challenge in time-varying channels. In this work, we address the effect of null space variations on the OGI mitigation. A measure for accuracy of null space estimates is proposed and bounds are derived on the residual average worst-case interference. We define the Null Space Update Rate as the inverse of the Null Space Coherence Time; i.e., the time it takes a null space estimate to become outdated relative to an interference threshold on the unintended receiver. In the case of Rayleigh fading channels, we derive a bound on the Null Space Coherence Time in closed form, and compare the performance of a given null space tracking algorithm against this bound. Both Monte Carlo simulations and the analytical bounds show that the worst-case interference levels are sensitive to null space variations, independent of the null space learning algorithm employed.