Purely a thought experiment:
Say I had a group of objects called "pins". I give them a property called "depth" and assign a value of N. Then I include the "pins" group inside another called "chip" with no custom depth. At output time, a final set of layers is created by parsing all of the objects and copying them to the output layers. Objects with a custom "depth" attribute like the "pins" group would be output to N-1 layers below the input layer, along with the corresponding output layer.
Or how about a set of named colors? EG, "registration" color outputs to every layer, etc.
I'm probably over-simplifying, but I face some similar issues when preparing screen prints: some objects need to exist on multiple screens (in exactly the same position). I end up doing a lot of back-and-forth between layers, pasting in place, spot colors (ie, I manage it all by hand - but it's not as bad as it sounds, I assume a PCB is going to be more painful though).
If this is way off base, feel free to ignore ;)
Chris
Chris