Even though the documentation clarifies that if an index has a collation, then a query will only use it if it specifies the same collation (but it doesn’t say that an index without collation cannot be used by a query with collation), I think it is not all good that an index without collation can be used in a query with collation (https://docs.mongodb.com/manual/reference/collation/).
It happened to me that, having a case-sensitive text field, I had to make a query sorting by that field but using the alphabetical order (case-insensitive). While I specified the collation in the query (which results in the collation being inserted into the command object), the result of the query is that the field values were sorted using binary comparison (case-sensitive). Analyzing the queryPlanner, I could see that the collation was not taken into account and the index without collation was used, causing the order to be defined by said index. If I also apply a filter on this field in the query to search for a word with characters similar to an existing value but not identical (differences in diacritics), the comparison correctly applies the specified collation and returns a correct result. In other words, for the filtering process the collation is applied correctly but for the sorting process the index is used, ignoring the specified collation and generating a wrong result.
That being said, if the collation is applied at the cursor level, the sorting is done correctly and the result is as expected.
PS: sorry for my english, I’m not native and did my best with google translate.