cache invalidation, when does it break

hey all, i've been thinking a lot about cache invalidation lately, especially in distributed systems. we often talk about strategies like ttl, write-through, or even write-behind, but what are the scenarios where these fail or become extremely complex to manage correctly. i'm curious to hear about real-world examples where cache invalidation caused major headaches and how teams eventually solved it. are there patterns or specific architectures that inherently make invalidation harder