The fast/history/history_reload.html test has a potential race condition caused by higher-resolution timers. I don't know exactly what threshold it takes to expose the problem, but it's worth removing anyway.
Each time the subframe loads, the onclose handler calls layoutTestController.notifyDone via setTimeout(0). If that gets fired before the history.go() completes, the test appears to have failed.
One solution is to always use two passes in the test, and not complete until the second one.
Created attachment 24322 [details]
No change to expected results.
Landed in r37657.