"I don't know what the reason for the 4-digit limit was," I guessed it would avoid building a full radix tree where a given hash was likely to be ambiguous, but maybe I'm wrong since it seems clear that linear scan in Python wouldn't be faster than building radix tree in C.