We allow querying tags and cdstubs by (row) ids, but we never actually use those IDs publicly that I can see. As such, these can't be used by anyone other than by entering random numbers and seeing what comes out, or by using IDs from a database dump or the like (in which case they probably do not need to query the search server to begin with).

These two already have other unique identifiers we do expose and should be using as such for the search at least: for cdstub we expose unique discid, for tag unique name.