Now, I'll concede out front that I'm not a member of a graduate admissions committee. Nor have I been admitted to a physics graduate program. But it seems to me that some of the information here is misleading.
As for taking the GRE twice, I don't see how that could possibly hurt your chances of admission (assuming you imporove your score the second time around). And why in the world would an admissions committee take the AVERAGE of your scores? That doesn't make any sense at all, since your most recent score indicates your present knowledge and ability in solving physics problems. This is a SUBJECT test, after all, so a good score means you know your stuff (i.e., it's cumulative).
It is better to explain your one score in the personal statement than taking it multiple times. This isn't like the SAT anymore. The top top physics schools are less concerned with your GPA and GRE and are looking for the exceptional research-driven people who can win them nobel prizes. The GRE and gpa are merely used to see if their applicants will pass their qualifications.
I agree with the last sentence here, but wholeheartedly dispute the recommendation of "explaining" poor scres (or GPA for that matter). Doing that will just make you sound like a whiner; let them make their own judgements about your record. However, it is certainly true that the subject GRE is viewed as an indicator of future success with quals. That's why it would make some sense to study and take it again if you get a low score the first time around.
As I understand it, attrition is one of the biggest problems facing graduate programs, particularly in physics and astronomy. The most important job the admissions committee has is admitting individuals who are likely to succeed in their program. In pursuit of this goal, the Physics GRE test is just one indicator of an applicant's potential for success. So they're not usually "looking for ... people who can win them Nobel prizes." Rather, they are looking for somebody whose record (GPA, general & subject scores, experience ... the whole bag) demonstrates their potential for successfully completing grad school and becoming a productive member of the research community. You don't need to look like Nobel material. Just "package" yourself in the right way so they know you've got the brains and the guts to finish what you start.
Here are some interesting articles and abstracts on this topic (many related to the gender problem):