The Courant–Friedrichs–Lewy (CFL) condition (C) is used to select a time step size to ensure convergence/numerical stability in solving discretised PDEs. Basically, the idea is to ensure that time step is small enough to capture any inter-grid points dynamics. Generally it is expressed as : u x sum[(del t)/(del X)] < C_max (usually 1.0).
To find out CFL for IMPES method specifically, the following paper should be helpful.
Study and Approximation of IMPES Stability: the CFL Criteria
If you're considering the stability of the IMPES method the classic reservoir simulation textbook by Aziz and Settari is well worth looking at. The relationship required between del X and del t for stability is thoroughly presented there.