I don’t feel it should be necessary for all identification tips to have a published citation attached. This would exclude all identification tips sitting in the heads of expert iNaturalist users. I can see that a citation gives an element of reliability, but published keys have plenty of errors or couplets which are at best unreliable. If tips were to require a citation, many skilled identifiers won’t get around to writing a published note on everything they know, and when they do it could be a year or more before the article is published.
The trouble with giving id tips under an individual species, as suggested above, is you generally need tips to separate two species. Saying how to identify species A requires you to guess all the other species someone could confuse it with. Better to invite tips on how to separate A from B. Then there is the question of how to index the tips.
For identification tips, you don’t necessarily need a key to a whole genus, just to two or three difficult species. And the difficult species may not be in the same genus. E.g., the fly you are most likely to confuse with Phaonia pallida in Britain is Thricops diaphanus.
There will remain the issues that tips may simply be wrong, and that they may only work in certain parts of the world. So I suggest a tip needs to have a field describing the geographical area in which it is valid; and there should be a voting system rather like the current RG system which allows iNatters to agree or down-vote it, preferably backed up by comments.