Varying degrees of SU’s within a hierarchy

[KirstenBirchHaakansson]

In the upper hierarchy the number of SU’s is fixed by the hierarchy, but in SA and the lower hierarchy the number of SU’s can vary between samples within a hierarchy.

I don’t see how this would work with estimators where number of SU’s is declared in the beginning. It could work if we scale to the level 0, then it easy to add it to the 'point estimate', but how would the variance be handled?

I can see it working with more generalized estimators, where these just pick up the level of SU’s served (no matter the number of servings), but I’m uncertain if I overlook something crucial.