cache invalidation strategies get complicated when you have a lot of moving parts, like multiple data sources or users

we've been using Redis for caching API responses and some computed data for a while now, and it's been great for performance. our current invalidation is pretty simple, expire keys after a set TTL. but as our application grows and data gets more dynamic, I'm wondering when this simple TTL approach starts to fall short. are there specific scenarios or data patterns where you've had to move to more sophisticated invalidation techniques, like event-driven updates or manual cache clearing? what are the trade-offs you've seen