@User, @User: I fixed the problem that I saw in netgen during symmetry breaking. Only. . . that was only a small part of the total time taken up by the symmetry breaking code. Most of the time is taken up by a method netgen uses for breaking symmetry by matching one (arbitrary) pair of symmetric partitions and then re-iterating the matching algorithm to convergence. If there are thousands of identical components like decap or fill cells, then it iterates thousands of times. It is not clear to me that there is any reason for matching one pair at a time. I tried matching everything in the symmetric groups all at the same time and then iterating the matching algorithm once. Seems to work okay. So I think I have solved the problem. I would be much more wary of modifying a bit of established code like that except for the fact that I have never seen a circuit that got to the stage of saying "matched with X symmetries" that wasn't LVS clean.