By going from charge distribution 0 to charge distribution A, an additional negative charge is added. My guess would be that if the negative charge epsilon was extracted from the electrons (so that they ended up with charges e – epsilon/2) would in total increase the energy of the system. Another thing that I noted was that the energy between the additional charge and the two electrons was not taken into account (I am aware that in total this contribution would be smaller than the contribution of the interaction between the positive charges and the additional charge and can thus not be used to explain the paradox).