I don’t think this would be a very useful feature - I think any outcome that you (general you!) might want to achieve using a tentative ID can already be achieved with existing features.
An ordinary ID is fine if you are ~fairly sure, even if not certain - it’s called “suggest an ID”, after all.
A higher-level ID (genus, family, whatever) will get the observation into the right filters for experts to find, without putting it into the position of “a single agree will make it research grade”.
A comment explaining your possible ID and any doubts will invite useful responses and contribute to others’ IDs too.
Setting “the community ID can still be improved” will prevent the observation going to research grade until you or someone else sets it back again.
I think one or more of those in combination can always achieve the effect of a ‘tentative’ ID, without needing to add complication to the system.
They can do that already by searching for observations with normal IDs, which will provide a much more reliable source of information. And trawling through higher ranks is often a good way to assess how difficult it may be to make lower level IDs. If I’m unfamiliar with a taxon and see that the top identifiers are consistently only making IDs at a higher level, it should give me a strong hint that I need to be more cautious.
Every Community ID is a work in progress. Research Grade is a temporary status indicator - not a fixed end point. It does not imply that any form of verification has taken place. Any responsible consumer of iNat’s observations will know that, and should either treat them as unconfirmed, or post-process them using their own verification system. As an example of the latter, iRecord automatically imports all of iNat’s RG observations from the UK. They have a pool of volunteer experts who independently assess them, and any verified observations are then passed on to the various UK recording schemes. The RG status is actually largely irrelevant - it’s mostly just a convenient way to filter out the low-quality observations and keep the dataset to a more manageable size.
It may be that this wouldn’t be a useful feature for most, but personally, I would find it useful - and this is a good way of finding out how many others would as well. Your points are well made, but for me, don’t really fill all the gaps.
@bazwal talks about Community IDs being a work in progress, which is correct in most cases. However, for some observations, there aren’t enough knowledgeable people to turn around an incorrect ID, which could stick for years. An ordinary ID, with one agree, can set an observation incorrect for a long time, longer than a “suggestion” should.
I’ve seen some observations where I have a choice between two species that are in different orders. I’m leaning towards one of those, but I’m not confident enough to make an ID. This means that I can’t confidently make an ID at anything better than Class. That particular class is a dog’s breakfast (Ascidiacea), and it would be difficult for an expert to trawl through all of it.
I absolutely agree in adding comments where you’re not sure of your ID. I have seen mine frequently ignored though.
I’d try to avoid the “Community ID can still be improved” where possible. Most likely people don’t know how to use it and would think something’s wrong with the observation.