Unfortunately, I don’t think it’s possible to separate these issues. Postal cities are just about the only reason why a subset of non-administrative place names would be relevant to geocoding in the U.S. Map users don’t care about a grand unified theory of place when it comes to describing locations or searching for places. They already have a mental model of place that they interact with every day. For users here, mixing formal and informal place names in the same hierarchy is a bug, not a feature.
Of course, a geocoder is welcome to do things differently. For that matter, a renderer is welcome to take artistic license and devise its own symbology that bucks real-world conventions in search of something more enlightened. But the OSM community also has an interest in the most prominent OSM-based geocoder eventually improving to meet normal user expectations, without turning every search into a human geography lesson. I’m not the only one who started out in OSM by incorrectly deleting place points or conflating them with administrative areas just so they wouldn’t pollute search results from Gazetteer and later Nominatim, before being chided for fudging the data.
We’ll keep correcting this sort of confusion when we see it. We’ll keep reminding people that it’s hard to make a single piece of software that works perfectly across such a diverse world. I think all of us look forward to Nominatim improving over time. We just don’t think the CDP boundaries can further that goal.
Dual tagging implies that the two features have the same spatial extent. Earlier, I included some representative examples demonstrating that CDPs generally do not have the same extent as their namesakes, especially at an address level. A better analogy is that we don’t tag the church grounds as if it is a parish boundary.
And they’re right! Ordinary people in the UK don’t use boundaries, place names, and addresses the same way as ordinary people in the U.S. use boundaries, place names, and addresses. In suburbia, most of the unincorporated place names that came in from GNIS are old-fashioned compared to named landuse areas, especially in states that have admin_level=7 boundaries. Many of these place names are still marginally useful as reference points on the map but not in search.
To be sure, the administrative hierarchy can sometimes be too rigid, surprising users who don’t know the local geography very well. I attended a school on the other end of Finneytown, literally steps away from Cincinnati, but would often describe it as being “in Cincinnati”. Per Postel’s law, some tolerance could be helpful when interpreting queries, especially in and around cities that have very irregular boundaries. But this probably looks very different than trying to treat informal places like formal places.