Well, as far as YSO is an ontological database, and accordingly its entries are supposed to be for places and not just place names, then we should be able to tell which place of the same name is its each entry for (former or current municipality, settlement or municipality, settlement A or settlement B). Many YSO place entries in that regard seem inadequate. For instance, I'm not sure based on what you decided that Sangaste entry is for a settlement, as YSO entry also mentions former municipality (kunta). Finnish-language Wikipedia article being mirror in that YSO entry also doesn't help, as that article confusingly mixes settlement and former municipality, while settlement and municipality co-existed, one didn't become another, and it's unclear which data is for which in that article. So, as far as I can see the distinct-values constraint is correct, and the problem is purely on YSO's side.
Currently there constraint violations relating to 54 YSO entries of Estonian places alone. Mostly these entries mix settlement and former municipality, some also mix different municipalities or neighbouring settlements. There are no actual duplicates on Wikidata, though the constraint confusingly may suggest otherwise. Due to current links between Wikidata and YSO, data about municipality may easily get used for settlement, or vice versa. Hence I believe it's better not link these YSO entries in their current form, to avoid further confusion.