It's a code quality problem. For people like you who haven't taken a peek at the code (which is obvious since you say it's only a change in two places) and don't have to work with it, it could perhaps appear as a phantom problem, but for maintainers it's a real one.
Anyway, I didn't bring that up as a reason for not doing it, only as an explanation why it hasn't been done already. rsearch() will probably happen eventually. Most likely it just won't do all the optimizations search() is doing.
/.../ the reverse takes little time compared to the search.
Right.. If that were true, it'd be a very bad outcome indeed for all Hubbes efforts optimizing that beast.