You don’t need a pure ethnonational country, though that can cut down on certain bullshit like what’s happening between the Buddhists and the Muslims in Myanmar. (Hint: It isn’t solely a story of purely poor, innocent Muslims, but one of those cases of cyclical retaliatory ethnic violence.)
The chief question is, are you willing to do what it takes to make that cosmopolitan polity not descend into retaliatory ethnic violence, potential ethnic predation, and ethnicity-aligned political parties?
Like, if you’re not willing to do that, then what you get is ethnic violence.
If you look over at what’s happening in the British cities, with the child sex trafficking, and grooming, and so on, and not only are you not willing to slam down the iron hand of the state to stop it dead in its tracks, but you won’t even stop them from marrying their cousins at rates way above what is normal or even healthy, then you don’t really have what it takes to make a cosmopolitan polity work.
And if you don’t have what it takes to make a cosmopolitan polity work, then an ethnic polity is a safer choice.
This is somewhat disguised by the fact that not all cultures are equally destabilizing. You can pretend, for a while, if the underlying conditions are right, and succeed by accident.
Additionally, cosmopolitan vs ethnonationalism is a continuum, not a binary. Well actually it’s a multidimensional space, not a continuum. But you get the idea.
Anything that promotes cultural unity over cultural diversity, even shared experiences, heads in the direction of ethnonationalism, even if it may be far from reaching it. Having one shared language, or having a required period of civil or military service, head in that direction. Even soft unifying forces such as television can count, without any force of arms in backing.
Cultures have more than one layer. American food court multiculturalism, in which deep differences are stripped away while surface differences, like cooking, remain, moves positively along the axis of ethnonationalism, even though many people will object that “American” is not ethnicity.
A theoretical total cosmopolity, in its maximal diversity, has one language for each inhabitant. A theoretical total ethnopolity borders on hivemind.