when do you add a distributed tracing system i mean its prob a good idea but when do you actually do it

Hey everyone, I've been struggling with observability lately, especially when it comes to tracking requests across our microservices. Debugging issues is getting tough because matching up logs from different services takes forever. We're thinking about using something like Jaeger or Zipkin. What's your experience been like? When did you decide it was worth the hassle of setting up and maintaining a distributed tracing system? Were there certain metrics or problems that made you realize it was time?