cache invalidation strategies, when does it get tricky?

Hey everyone, I've been thinking about caching lately, especially when things get complicated. We all know caching is great for performance, but the real pain point often comes with invalidation. When do you find your cache invalidation strategies start to break down or become overly complex? Are there specific scenarios, like highly dynamic data or complex interdependencies, where a simple TTL just doesn't cut it anymore? What approaches have you found effective for managing cache invalidation in those tricky situations?