The performance of distributed constant false alarm rate (CFAR) detection using an improved version of the variability index (VI) CFAR as local detectors is analysed. The VI-CFAR which is a combination of the cell-averaging (CA) CFAR, smallest-of (SO) CFAR and the greatest-of (GO) CFAR is robust in non-homogeneous environments. In this work, we improve its robustness by making the VI-CFAR switch to the ordered statistics (OS) CFAR in multiple target situations where the interfering targets are present in both the halves of the reference window. Simulations for the "AND", "OR" and the "MAJORITY" fusion rules are conducted to assess the detection performance and the results are presented and discussed.