Cache invalidation: TTL or event-driven

hey folks, ive been wrestling with cache invalidation strategies lately. weve been relying heavily on ttls, but it feels a bit blunt. sometimes data gets stale for longer than necessary, and other times, if ttls are too short, were hammering the database. im curious to hear about your experiences with event-driven invalidation. do you find it more complex to implement but ultimately more efficient? or is ttl still the go-to for most scenarios? what are the gotchas i should be aware of if i move towards an event-driven approach