I understand the use case of distinguishing between different thematic layers from GIS perspective. But, due to the data model of OSM, the only possibility of type distinction of polygons is through their attributes, isn't it? A GIS application will always have to apply a filter on these attributes in order to separate polygonal features into separate thematic "layers". An then it should not make a difference if one polygon describes two different types, or, if two geometrically similar polygons describe one type each. Though, I see a disadvantage of the latter variant, where housekeeping of (nearly) duplicate geometries, in order to hard-codedly separate the attributive meaning, can lead to an overhead from data/geometry maintenance perspective.

If both polygons significantly differed in their geometrical shape, I would for sure see it valid to model them as separate geometries. But in this very case I carefully compared the data and, also with local knowledge of having been there, came to the conclusion that my change would be beneficial when trying to avoid data duplication. Also, as far as I remember, my result rendered more conveniently in OSM-based mobile mapping apps.

I am open for more thoughts from your side. I will also re-evaluate my change and can revert it if it breaks conventions or poses other problems which I have not
been aware of.