Added the failing expected results in http://trac.webkit.org/changeset/83738
The XP bots just went back to producing the old results in <http://build.webkit.org/builders/Windows%20XP%20Debug%20%28Tests%29/builds/27622>.
I suspect this test's output is sensitive to whether or not some earlier test has run. As tests get added/removed before this one in the list, the point at which we relaunch DRT changes. If the trouble-making test happens to run in the same DRT instance as this one, we fail. If not, we pass.
We should probably just skip this test. My guess is that bug 55329 is a similar issue.
Skipped the test
Committed r83965: <http://trac.webkit.org/changeset/83965>
This does not fail on any test system.