Historically this is how the number of "not run" tests is calculated.

        int p = stats[Status.PASSED];
        int f = stats[Status.FAILED];
        int e = stats[Status.ERROR];
        int nr = stats[Status.NOT_RUN] =
            harness.getTestsFoundCount() - p - f - e;

However it is possible that in special circumstances the getTestsFoundCount() number could be lower than amount of passed tests - if tests are added dynamically.

This could lead to negative number of not-run tests and could confuse stats formatter