Commercial Macsyma has several choices ...
to solve
solve([(x-a)^2 + (y-b)^2 = r1^2,(x-c)^2 + (y-d)^2 = r2^2], [x,y]
on my home computer ...
using triangularize: 78 ms
using grobner: 516 ms
using resultant 25,484 ms. and then returned [].
So it seems that using the triangularization method is the winner.
A debugged resultant presumably would take less time than the existing one.
I don't know how it would compare with the others, though.
RJF