You very well captured the intention of this (or at least what I think is the intention). I mentioned above that in highly abstracted systems it is not at all u

You very well captured the intention of this (or at least what I think is the intention). I mentioned above that in highly abstracted systems it is not at all unusual to spend more time with inefficiencies rather than doing real work: You noticed here that it is also hard to tell in a high abstraction layer which operations might be fast and which may be slow. Often this is even found in seemingly simple places, e.g. traversing a List<> in Java can have vastly different performance characteristics (LinkedList vs. ArrayList).