Interesting read. I've glanced over sections and read others.
I've been involved in the use of some of these frameworks. They are good reference frameworks but they are most often overly heavy to get started and/or the implementation is most often not supporting what the marketing material suggests.
Good integration projects start from limited partners with limited datasets exchanging with each other knowing they want to exchange with more partners in the future. Limited can be two to seven and they can even be in the same organisation. Perhaps it works with more, I have not been involved in that. They contain only the necessary roles and know how they'll expand. These frameworks are great to get the expansion options in place. However, to get something going we've seen that a cheap and structurally correct solution to data sharing gets the ball running more quickly and for longer, then an expensive and heavy project which needs all sorts of approvals (even if it would be superior).
Projects that start with the weight of all the roles filled in by different actors need substantial funding to get started and I don't think they generally last after funding dried up. Projects faking their way through the extra roles and providing the most limited interpretation of some of these roles with the project having some (even vague) shared benefit seem to survive past the grant stage.
Get some basics right and get going is my advice. Identifiers, shared and extensible data models, and a clear way of sharing data.
I like these frameworks, they help set the stage, but my gosh do they see problems where there are none when a project is just getting started.