so, when do we actually need a distributed tracing system

hey everyone, i was thinking a lot about observability lately, especially in microservice architectures. we've got logging and metrics covered, but distributed tracing feels like a whole other beast. when do you all find it necessary? is it only for massive systems, or are there smaller setups where it can provide value? what's your tipping point for introducing something like Jaeger or Zipkin? cheers