According to my compilers class professor there’s a broad literature of optimizing the “giant table with all signatures” approach with heuristics for saving spa

According to my compilers class professor there’s a broad literature of optimizing the “giant table with all signatures” approach with heuristics for saving space by rearranging and merging the tables of different classes into the same space or re-using offsets across classes to make tables smaller. But the general problem is NP-complete so can only be solved heuristically. A SAT solver should actually be able to handle this fairly easily, because in practice, you don't get long complex chains of overlap between the interfaces implemented by common classes. If someone feeds your compiler an adversarial input you can just fall back to the naive implementation.