@User: So it looks like the "conb" cells have the opposite effect of the antenna cells, creating lots of partitions with just a few cells in each. . . Probably this can be fixed by running all of the symmetry breaking on all of the partitions at once, instead of just within a partition, which is what my last fix did. Also: This may be the more important case, since it cannot be solved by combining cells in parallel, which would probably have fixed the case with antenna and fill cells without the need to modify the code.