The semantics of IfcRelNests as opposed to IfcRelAggregates was obviously changed in IFC4. While in IFC2x3 the difference was whether the parts and the whole of a composition are of the same or different type, in IFC4 the difference is whether the parts have an inherent order or not. I don't want to debate if a fundamental change of semantics without changing the name of an entity type is appropriate. I just want to point out that the specification lags a bit behind that change: the supertype IfcRelDecomposes still mentions that "Decompositions may be constrained by requiring both, the whole and its parts, to be of the same type - thus establishing a nesting relationship." This will cause confusion.
This was initally reported in the buildingSMART forums.