Recently some nodeagent unit tests have started failing, but this only happens every once in a while. Currently we haven't found a way to replicate this except running the tests multiple times.