The duration time is much larger than x-envoy-upstream-service-time. What are the possible reasons?

In some case(i don’t know case is what) , duration is much larger than x-envoy-upstream-service-time

like this log

 HTTP/1.1" 200 - via_upstream - "-" 0 13297 83 42

It seems to be random, and I don’t know how to reproduce it or how to debug it.

Does anyone know the possible reason?