Minor DocumentModel has a notion of prefetch but it's not useful, as actual prefetch is done at a lower level anyway, and the implementation has complexity that can be removed.
There is a (small) side effect to this change. We have two ways of representing simple lists of scalars in Nuxeo: the recommended way, xs:simpleType, is represented as an array in Java (a subclass of Object[]). The alternative way is to use a xs:complexType with just one element, and this is represented as a List in Java. For this alternative way, previously if the field was configured as prefetched, the Java code was incorrectly normalizing it to an array instead of a List. As there is no more prefetch this doesn't happen anymore.
Another side effect is that code that expected DocumentModel.getTitle() to work on a detached document without the dublincore schema will now fail. Such a document must be fully detached with DocumentModel.detach(true) or the dublincore referenced before the document is detached.
Note that prefetch still exists at a low level in VCS.
