I guess I understand now the process...
and I guess that was my first and last translation.

This process is not like modern engineering, where you get early feedback.
Better process would be that translations are generally approved
and new translations are blocked, if someone of the first users complains about something in the translation.

Thanks for the hint. Change is done.

How long does it typically take until a translation is approved?

Thanks for recommendation, I added the @DisplayName annotation.

Check row can be done easier with: for (int[] row : sudoku)

My algorithmn fails at "moreDifficultTests" with 23 mines.
My solution is not using the knowledge about having 23 mines in total in this field
and I did not implement a back-tracking algorithmn.

I guess this is not solvable without using back-tracking and the knowledge about 23 mines.
Is that the case?

In Java version all assertEquals have wrong ordered parameters. The explaining message is set as expected so the tests can not pass.

For Kotlin in assertEquals the parameters are switched.
It should be first expected and then actual.

But then the information "n <= 1500" is irrelevant. I would work also for maximum integer.

Kata descrition says "1 <= n <= 1500", but this is not covered by its test cases.
Tests for numbers outside this range should be added.