Traditional TCP-Reno like congestion control protocols exhibit poor performance when deployed in fast or very large network paths. Delay based congestion avoidance mechanisms (DCA), like FAST-TCP, get much higher performance, in the same circumstances. However, when mixed with TCP-Reno or alike traffic they are unable to attain their fair share of bandwidth. In this paper we present a new mechanism that can indirectly detect the present of non DCA-friendly traffic that can be used by new DCA algorithms to auto-tune themselves with more aggressive parameters to achieve their fair share.