regarding ‘I do agree it’s more helpful to help with a backlog of Unknowns by starting with older observations, not the most recent ones.’, I have to disagree on this.
If a new observation arrives with an Unknown label, the quicker you can get some or any classification onto it, the greater the chance of folks specifically doing reviews in that area will see in it their passes through the data.
In my experience, adding a coarse ID to a new record has a much greater chance of getting review than doing it to older ones. It’s not that older ones don’t gt any results, just lower rates.