Magnetic Flux Leakage (MFL) method is widely used in non-destructive testing of ferromagnetic specimens such as steel pipes, ropes and rails. Inspection velocity is a critical factor in online rail track evaluation due to specific inspection conditions. As the speed increases the distribution of magnetic induction inside rail becomes inhomogeneous primarily under the influence of eddy currents. This complicates MFL signals interpretation and reduces the ability to detect deep subsurface defects in the rail head. In this paper we explore the speed limitations of the traditional MFL rail inspection and contemplate the ways to overcome the said limitations. To investigate the dependence of flux leakage data on inspection velocity two different methods were applied. The first one is 3D computer simulation of interaction between the fixed rail and the moving magnetizing system. The second method is the set of field measurements in which an experimental setup consisted of electromagnets and sensors moved along rail with artificially made defects. The results of both methods indicate to detect defects located in the center of the rail head at speeds over 80 km/h the distance between magnetizing system poles should exceed conventional 3 m or some alternatives should be employed such as MFL remote field analysis.