Speaking from a perspective of statistics, your null hypothesis would be that your system has an existing error. The fun begins when you believe that test 123 detects error 123.
There is a good explanation at http://en.wikipedia.org/wiki/Type_I_and_type_II_errors, which has several phrases of various registers.
Type I error - error of the first kind - false positive - crying wolf
Type II error - error of the second kind - false negative - wolf present but failed to see
The latter set is from "The Boy Who Cried Wolf," one of Aesop's Fables.
You correctly applied the term false positive, which means: although the test says there's a problem, there really isn't one.
I think that if you said test 123 was occasionally "crying wolf," then even pointed-haired bosses would understand (perhaps better than "Type I error" or "false positive").
From a system integration point of view, it would be best to fix the test so that it reliably detects the problem.