The way I see it, there are 3 options for how to deal with “can’t be ID’d to species” observations: either you call them all casual, you call them all RG, or you pick some arbitrary taxonomic cutoff for which ones are casual and which are RG.
If you call them all casual, you lose a lot of valuable data being exported for groups where species-IDs are impossible from photos. If you call them all RG, you end up exporting blurry blobs of goo to GBIF as “life”, which doesn’t help anyone. So I think a cutoff somewhere to maximize useful data exports while minimizing useless data exports is the way to go. Obviously the very nature of taxonomy is such that wherever you make that cutoff is going to be sort of arbitrary, and you’ll always have cases right near the line where someone says “this specific set of observations isn’t being treated the way I’d like”. But treating the RG-cutoff for every lineage on iNat differently seems like it would create more confusion and difficulty than it’s worth. So are the cutoffs for RG/Casual arbitrary? Sure. Are they still helpful in the grand scheme of things? I’d say so. “All models are wrong; some are useful.”
I think the solution is just for researchers in those specific taxa to be aware of the fact that some generic-ID observations exist on iNat and aren’t being exported. If someone works with a taxon where lots of observations are going to sit perpetually at order or family level, they’re probably well aware of this, and with a few clicks, they can bring up the full list of all those “Casual” generically ID’d observations to do whatever they want with them. I don’t think there’s any need to overhaul the system and go to an “export everything” or “export nothing above species” model, nor would it be practical to get experts in every taxon of life to request a different taxonomic cutoff for exporting data in their taxon of study.
I think the current system is a good compromise, at least until some billionaire genetic purist buys up iNat, re-defines all taxa at all levels to be monophyletic clades defined exclusively based on branch lengths, and implements a cutoff for what gets exported to GBIF based on the percent of nucleotides that the CID is in agreement on. /s