The article is interesting. It highlights exactly some situations encountered when deploying. Instead of executing one big script which does it all, sometimes, some breakups had to be done.
I would say point 4.2 is probably the one which was making it more difficult to deploy seemlessly.