I’d welcome this change. The subarea relation members create an unnecessary burden on both mappers and data consumers. Moreover, this tree structure misleads both mappers and data consumers into incorrect topological assumptions about boundaries.
That said, I’d like to make a special request to leave the subareas of Indian reservation boundaries untouched for the time being. I’m no fan of subareas in that context either, but we have no suitable replacement yet, so deleting the relation memberships would lose valuable data that isn’t very easy to come by. I’d hope that the even more exceptional presence of those subareas would push us closer to a consensus on an alternative solution.
Apart from that, if you come across any subareas that for some reason don’t overlap their parent boundaries, please exercise care, in case the subarea’s or parent boundary’s geometry needs an update.
Once this change is complete, we should update the wiki to explain that the subarea role is not to be found within the U.S. and refer to alternative tagging.