I agree with most of what you wrote, but I felt the need to clarify that, even in the present day, a great many legal boundary descriptions still consist of a long list of references to physical objects on or in the ground. This may not be the case anymore in Germany, but countries with British common law influence have some very different traditions. Here’s an example from the U.S. of what metes and bounds looks like in official records (physical objects emphasized):
Beginning at a spike in the center of Hawley Road, the said spike being South 4 degrees 00 minutes West 1663.50 feet from a point at the intersection of the centerline of Jackson Pike with the centerline of Hawley Road; thence from said beginning point and leaving Hawley Road South 80 degrees 51 minutes East 212.65 feet to a stake; thence South 8 degrees 45 minutes West 185 feet to a tree and stake in a fence; thence with said fence North 73 degrees 25 minutes West 201.25 feet to a spike in the center of Hawley Road; thence with the center of said road North 4 degrees 00 minutes East 159.62 feet to the beginning, containing 0.81 acres of land, subject to all legal highways and any other easements of record.
This example is just for a parcel, but everything else can be defined in this same manner. There’s a whole industry of surveyors who make a living triangulating and digging up these objects, and the law prohibits unlicensed people from tampering with them. Boundary monuments conform to strict standards, and there’s a provision for reference monuments when it isn’t possible to place a monument at the intended location for some reason. We’ve imported lots of boundaries from government GIS layers, but that data is intended to be for reference only; it explicitly isn’t authoritative:
Of the 50 U.S. states, only three are defined solely in terms of coordinates. Wikipedia has a nice little footnote about Colorado that illustrates your point:
The official Four Corners Monument is located at 36°59’56.31608″N, 109°2’42.62075"W, 574 feet (175 m) southeast of the 37°N, 109°02′48″W location Congress originally designated.
But mixed definitions are far more common. To pick on Ohio once again, its land boundaries are defined by monuments, but the northern boundary is defined by coordinates in an international treaty, while the southern boundary has been lost to time and is thus defined by a GIS layer under a court-ordered settlement.
At least someone knows where that boundary is. The Michigan–Indiana state line, on the other hand, is still lost: it was staked out using wooden posts that have long since rotted, and the two states haven’t gotten around to resurveying it yet.
Whether we map the boundaries based on published legal texts or observed monuments and markers, there are many potential sources of error that all but disqualify OSM for critical use cases. The coordinates are usually defined in an old datum; many of the boundaries in OSM have been incorrectly transformed. Even if we see a monument or survey marker in aerial imagery, that imagery can be several meters off. As I see it, the only benefit to connecting the boundary to the marker is that we can more clearly express what the marker is for. It’s an extra level of micromapping. 