I've been able to get this test to fail consistently about 1/3 of the time on my local cluster. I checked against branch-2.7 and trunk and it failed the same in both. Since it's checking time on metrics, it will fail if the test runs too quickly, which is something that does not often happen on Jenkins. This would explain why we don't see it fail on there. However, without any load on my machine, I can get frequent failures. If I increase the load on my machine, then the test does not fail.
The interesting thing is that GetEditAvgTime == 0.0 in the test runs where it fails, and == 1.0 in the test runs when it succeeds. It's being treated as a double, but I only ever see it manifest as an integer. My guess is that the metrics are somewhere truncating the value and so when the time is between 0 and 1 it just truncates the decimal place, thus making it 0.