It fails on creating an Array size of Infinity.
I'm not sure when the test started failing (sometime in November 2012).
Oddly, the proper exception is indeed being thrown. The test does not properly catch the exception however.
The previous test in this file new Array(-1) also throws a RangeError but the error is indeed properly caught.
In the working case the error is thrown from JSArray.cpp
In the broken case it is thrown from ArrayConstructor.cpp
I'm skipping the test for now to get the bots greener. This seems like it could be some sort of Windows DRT limitation as well, who knows.