with the M button and in LUA.
Based on what I know about the Rosetta engine used by Foldit, my guess is that this is a technical limitation. To see why, look at the first few lines of the file used to define a cysteine residue involved in a disulfide bond (CYD.params):
IO_STRING CYS C
Basically, the issue is that one can only mutate to residues that are of the same variant type. This makes sense, because the disulfide bond would otherwise be lost. But since no other amino acid can form disulfide bonds, one would not be able to mutate the cysteine to anything else. A possible technical workaround might involve manually tweaking the side chain configurations so that the disulfide bond disappears; and if that isn't enough, one might have to move the backbone pieces further apart so that a disulfide bond is considered chemically impossible by the program.
Tweaking the sidechain is enough, the cysteine is then mutable.
Is a good disulfide with very bad sidechain score mutable? It may be that disulfide + other scores > any other aa score. If both are negative and it won't mutate, that is a bug.