It’s not just a problem of grafting to the nominate subgenus (which is clearly wrong). It’s also a problem when it grafts to the genus (which isn’t really wrong, it’s just too coarse). If the subgenus it belongs to exists in the iNat taxonomy, then grafting to the genus can cause unintentional disagreements in IDs.
As an example, let’s says species Foo bat is in subgenus Bar, but in the iNat taxonomy it’s grafted directly to Foo instead of to Bar.
On an observation, one user adds an ID of subgenus Bar, while another user adds what they think is a more specific ID: Foo bat. Both users intend to agree that it’s in subgenus Bar, but the system registers a disagreement because the system thinks Foo bat is not in Bar.
To my knowledge, there is no written statement requiring the locking of genera that have subgenera. Rather, it’s a practice that has arisen from curators deciding that they would rather be flagged right away when a species is added that might need a subgenus instead of allowing it to incorrectly graft only to be discovered and corrected later down the line.
I don’t think manually locking all genera with subgenera is the best plan. Personally, I would do it for the genera that get a lot of imports (e.g. Camponotus), and then ask the devs to add a check to the importer that would automatically prevent a graft to genus if the genus has subgenera.