From an architectural viewpoint, linked files are problematic. It's rare that a model is entirely self contained. Linking (XREFs) are commonplace and loading external files can be problematic. CAD models also gather 'junk' - symbols (blocks in AutoCAD parlance), which need purging from time to time.
One of the things it says is why Autodesk went to a subscription model. Another is why Autodesk deprecated AutoLisp. These on top of how little the economics of most architectural firms benefit from new features.
Autodesk are an absolute blight on the architecture and engineering worlds. While there is hand wringing going on around Apple/Google being investigated for antitrust, one of the biggest offenders goes unnoticed because it is 'niche'. Architectural firms don't make much money at the best of time, these parasites suck the life out of them.
Reconciling the MEP layouts tends to be manual, but it's just part of the job.
Automation may happen on large projects, but that's not most projects.
Is it worth solving? No.
The economics don't work out. The problem has been solvable for more than thirty years. The nature of the business relationships and the cost sensitivity of design fees and the disruption of changes to methods make it mostly a non-starter.
In mathematical terms it's in NP. A job shop scheduling problem where each workstation has no slack capacity.
There's no money in solving it. The limits on profits are almost entirely in the class of clients a firm has not in its work methods.