I see what you're saying and mostly agree, but... This could totally come up if you have a non factorising multivariate posterior distribution, and you want so

I see what you're saying and mostly agree, but... This could totally come up if you have a non factorising multivariate posterior distribution, and you want some sound way to summarise it when you can't reason from the marginals alone.