Compare the descriptions of File:Skulptur_Hand_und_Traube_548_bei_A-2170_Wetzelsdorf.jpg & File:Ingeborg Strobl - Mahnmal für verlorengegangene Artenvielfalt 1997 (1).jpg. While the first file shows the description: Skulptur "Hand und Traube" bei Wetzelsdorf, Gemeinde Poysdorf, Niederösterreich, Österreich, the second shows Dieses Foto zeigt das im digitalen Kulturgüterverzeichnis des Landes Niederösterreich (Österreich) unter der Nummer. (lang German selected). (you can use https://de.wikipedia.org/wiki/Liste_der_Kunstwerke_im_%C3%B6ffentlichen_Raum_in_Nieder%C3%B6sterreich/Weinviertel to run the mediaviewer)
I would expect that all the language specific parts of the description are shown (multiples possible) and that all non-language specific parts are shown too. Furthermore I would expect that templates are transcluded correctly (which means that that number in the second case must be visible).
It looks like that only the last matching paragraph is shown as the description in mediaviewer. If only one paragraph is shown, than the first one should be preferred.