We have noticed that changes(observer_dhcp_latency[1m]) sometimes goes to 0. During the build-up and tear-down of DHS19 it happened, most likely related to the network port flapping on switch reboot or equivalent.
Stdout / stderr did not show any issues, maybe a stuck poll?