actually on kubernetes, for the server at least, 1280 turned out to be too high. We had to bring it down to 1024, so that might be worth trying