Hot on the heels of me actually trying to do this (see bug 178207), it looks like there are a large number of things to investigate, and very few of them were the sort of tiny pixel-diff failures I expect from a compiler or architecture change.
Running the tests (using the -t Release_x64 arg), I got ~450 failures. Many of them were in the windows-directwrite suite, but the failures there looked like cases where we were seeing characters moved by multiple pixels (and the failures were not happing on the 32-bit build at the same revision).
There were a lot of other failures as well, though. Oddly, many of the shadow dom tests were failing, as if it wasn't enabled at all. There were failures in rubberbanding, and softwarecompositing, and editing as well.
I saw no obvious indications that there was anything wrong with run-webkit-tests or content_shell that would be causing these failures, unfortunately.
Someone else should probably attempt to reproduce my results, and then we should figure out a plan to triage and deal w/ these failures.
I unfortunately do not have much time to devote to this at the moment.