Tests are designed so that a score will always mean the same across tests. That is, a 32 will always mean a 32. While an individual may do better/worse on a certain test day due to certain topics showing up, the score itself is supposed to mean the same thing.
Percentiles, however, are not necessarily consistent across all testing years. The AAMC just reports the percentile of the PREVIOUS year's test takers who achieved your score. So while it shouldn't vary that much from year to year, it is conceivable that the percentile does shift (likely slowly getting lower, as in a 32 is currently in the low 80s percentile but will, over time, break into the high 70s as applicants get more competitive).
I think that answers your question. It's important to note also that the test is not curved, it is scaled. Because the AAMC uses test takers to establish validity of questions, they can simply design a test that should result in a similar distribution of scores as all previous MCATs. So, theoretically, a 32 ALWAYS means a 32 and will always mean a 32 (what changes, however, is how competitive a certain score is in terms of applying). It's basically 'curved' based on all examinations, there are no adjustments made to the scale for performance of a specific testing date. If a scale says "90% raw score = 40" and everyone shows up and gets 90% raw score.. everyone gets a 40. Likewise if everyone misses a certain question, etc.. because the AAMC already knows what to expect from the average test taker based on their experimental data. Scales do vary across testing dates.. slightly.. to adjust for small differences in difficulty. Again, this is established before the test is taken.
