If you re-read this thread, you may notice that in post #8, gneill said (paraphrasing), "with conducting spheres, it's complicated and not intuitive". That is an extremely strong hint that you cannot blindly apply the formula ##PE = k\frac{q_1 q_2}{r}## to the case of two charged conducting spheres.
What you could do would be to use the formula ##PE = k\frac{q_1 q_2}{r}## and integrate it over a distribution of charges on the two charged spheres, searching for the distribution that gives the lowest possible result. But that is more easily said than done.
Edit: There is a tricky piece in that process. We had assumed that you are concerned only with the potential energy of the two charged objects in relation to each other and not with the potential energy of each objects's charge distribution with itself. But the minimization process described above must take into account each object's own self potential energy. That means that the result depends on exactly what question you want answered.
Edit: Unless I am mistaken, it gets even worse in the case of more than two bodies because one might be faced with the possibility of multiple local minima. A naive optimization approach based on incremental relaxation may not lead to the global minimum.