The database is currently configured to use the collation Latin1_General_100_CI_AS_SC_UTF8 but that does not work for Arabic characters, and it has also issue with some Eastern European diacritics.
I completely lost track when attempting to get the relevant information out of the documentation provided by Microsoft; I also have to confess that I had no exposure to SQL Server before I got order to investigate why Arabian texts were converted from نورديا to ?????? when placed in the database and read again.
Another nasty thing is that "Słowiński, Tomasz" is converted to "Slowinski, Tomasz".
I checked https://stackoverflow.com/a/52480489/1554195 but I am unable to assess the value of that answer.
What is the meaning of the prefix N in T-SQL statements and when should I use it? does not answer this question by any means, because the SELECT statements are fix as part of an existing product and has to work with other RDBMS, too (at least with Oracle, PostgreSQL, DB2).
Edit: Perhaps it is important to note that the database is accessed through a JDBC driver from Java.
Edit 2: Further investigation showed that the collation Latin1_General_100_CI_AS_SC_UTF8 is really the correct one! But changing the collation for an existing database with existing tables with existing columns would not do the job! The columns will keep the original (wrong) collation, you have to alter each and every column to use the new collation …
That explains why you had issues to reproduce the problem: you created the database and/or table from scratch, immediately with the proper collation configured.
And I would like to put this edit as an answer, but because someone still thinks that my question is a duplicate to a completely unrelated question, I can't.