Is omega(3) really less than 1/2? It seems to hit 1/2 exactly at y[2]->Sqrt[15], y[3]->Sqrt[5/3], and the gradient is 0 there.