Browsing various PhD admissions programs, I consistently find that there is a higher expectation of international students on the physics GRE for admission. For example, at UT Austin, the average score on the physics GRE for students accepted for Fall 2011 was 907 for international students and 777 for US students.
Why is there such a difference in expectations?
(Some) US departments seek to maintain a critical mass of US students. (I expect that) such departments accept a critical mass of US students - the best that they can get - and then accept some number of international students to fill up their ranks - the best that they can get. Since international students generally score higher on standardized tests such as the GRE general/subject test (for all sorts of reasons), this leads to the accepted international students having higher test scores than the accepted US students.
A more general point, I believe, is that international students are often 'unknown quantities'. While a US student might have recommendation letters, research experience, etc. that gives an admissions committee a well-rounded perspective on them and could potentially make up for less-than-stellar test scores, this is not always possible for international students. So in some sense an international applicant has to have exceptional test scores to make up for being an unknown quantity in other respects. I believe this partly because international students who have been undergraduates in the US seem to fall into the `domestic students' box more so than the 'international students' box.
The physics GRE is not a very good test of one's knowledge of physics or ability to solve physics problems. Most physics homework questions students will encounter are quite a bit lengthier and more involved than the sort of question that appears on the GRE. What the GRE tests is the ability to solve lots of very simple problems very quickly. Students who haven't specifically practiced for that will often do poorly even if they know the material well. Physics departments in the US typically don't give students this sort of practice. In some other countries, it's much more standard for departments to encourage such practice and provide assistance with it (e.g. students might train on lots of old exams). So it's a mistake to interpret this statistic as implying that international students are better educated than US students.
Another issue that may play a role is that it is much harder for faculty to accurately evaluate international students than US students. The recommenders and their institutions are less likely to be known to the people evaluating, and there is generally less of a good match between expectations in the different systems.
Thus, it is often the case that an international student needs to be much more obviously excellent than a US student, in order to obtain admission to the same program, and this would be expected to be reflected in GRE scores as well.
First answer: education systems in the rest of the world include physics earlier in the curriculum than in the US, so one would naturally expect a student with more experience to score higher.
Second answer: Some countries have explicit "GRE training" to get their students into the highly-regarded US higher-education system (grad school). (this is an editorial answer - I have heard of such things but not in an official capacity).
Third answer: Before you get to graduate school, science/math education in the United States is terrible (googleable fact). If you held students to the same requirements, there would be zero US citizen graduate students in US graduate programs.
