So far, I see a lot of blanket assumptions being made in this discussion. In my experience, even a single specific boundary can be legally defined by a mix of coordinates, angles, and references to physical features. In the parts of the U.S. where I map, this is the norm for everything from property lines to city lines to state lines. Even time zones.
The norms around boundary definitions vary from country to country, even within a country. For example, here’s a map of Ohio by survey system:
I grew up in the southwestern corner of the state, close enough to see the river that forms the boundary between the Virginia Military Survey (in green), where boundaries continue to be defined in terms of traditional English metes and bounds, and the Symmes Purchase (in blue), which uses the modern Public Land Survey System to define boundaries. As a result, I have to be prepared for very different-looking boundaries on either side of the river – even in the same town.
I happen to agree that historic boundary markers tend to be somewhat irrelevant for modern-day boundaries, so they should be kept separate from the boundary relations. However, this is only because the markers have often been superseded by something else, whether a modern survey marker or a coordinate. If someone has gone through the trouble of mapping a modern survey marker that the legal definition literally refers to, then I see no problem with including it in the boundary relation. That said, that kind of detail is extremely rare around here. We still have hundreds of missing admin_level=7 boundaries to map in the first place.
There are some limitations to this approach. Since mappers have such an allergy to anything overlapping or connecting to roadways, it isn’t possible to attach things like parks and schoolgrounds to any pins embedded in pavement that represent the property line. Also, when a boundary description literally says “thence along the centerline of” a street, we can’t necessarily attach it to a highway=* way, because we might have split the road into a dual carriageway or offset the way for other reasons.
All told, I’d tolerate markers as part of boundaries in some cases, but I wouldn’t make it a requirement. OSM makes a lot of compromises that you wouldn’t see in a formal GIS parcel or boundary dataset.