One other thing that seems worth mentioning is that as soon as you
jump from relational to object-relational modelling is that the latter
is more rich and hence more complex than the former. Because
object-relational modelling is a much expanded semantic superset of
relational modelling, the antipatterns are a much expanded superset
there too. Additionally because the patterns are not yet well
understood, the whole area needs to be seen as somewhat immature.
That doesn't mean that the features are primarily useful as foot-guns,
but it does mean that productive use of features like this involves
careful documentation of patterns and antipatterns.
Best Wishes,
Chris Travers