When should we think about eventual consistency instead of strong consistency?

So, I've been thinking a bunch about consistency models lately, especially for distributed systems. We usually just go with strong consistency 'cause it's easier to wrap your head around. But, if you're running high-traffic apps spread all over the globe, that strong consistency can really hurt performance and availability. I'd love to hear what you all think. When has eventual consistency actually been the way to go for you? What tells you it's the right choice, and how do you handle the tricky parts that come with it?