a couple of things here.
first, not every observation is created and identified (correctly) at the same time. so if you’re trying to detect, it will often be faster to look for new identifications than it is to look for new observations.
second, because of what i noted above, there’s not always an easy way to determine which observations were added to such a project recently, even if you’re sorting by observation create date. that means that to find observations that are newly added to the project, you’d almost have to mark everything there as reviewed and then constantly check for unreviewed items (because you don’t have a way to sort by date of a particular relevant identification or by date added to a project).
with any detection approach, you’d probably have to wade through some hay to find a needle, but the kind of approach described here gives you a ton of hay to wade through i think.
i’m not questioning the need for detection. i just struggle to understand why the kind of workflow that would be implied by this proposal would be the most effective for detection.