It will not change the percentile rankings any more than any other new GRE test changes the percentile ranking (i.e. each year, the population of score used to calculate percentiles is updated because they add the most recent year's tests and removed the oldest year).
Remember that the scoring goes like:
Raw Score (out of 100) --> Scaled Score (out of 990) --> Percentile Ranking
The point of converting Raw Score to Scaled Score is to normalize between difficulty between test years or test versions within the same year. So in theory, everyone's raw scores after September 2017 will be a little higher than the raw scores of test takers prior to Sept 2017. However, the distribution of scaled scores should not be affected by this offset. That is, if everyone gets X percent more points than they did previously since they no longer lose 1/4 points for wrong guesses, this is similar to applying a constant in the raw score and that would be scaled out.
However, I can think of one reason why the scoring might be messed up. What if a large fraction of people didn't know about the new scoring changes. In the worst case scenario, 50% of test takers don't know the change and 50% of test takers do. Then, only half of the test takers will get a raw score "boost", resulting in them getting disproportionately higher Scaled Scores.
But I don't think this is a major concern. This change provides a slight advantage for September 2017 test takers that listened closely to the instructions but I would imagine the small change in scaled score is far less than the ETS published "uncertainty" in subject scores (close to 70-90 points in the 990 scale).
This change will also not affect the percentile ranks of anyone else who took this test prior to this change, for now. The Scaled Score-->Percentile Ranking table only gets updated once per year (July 1). All your scores reported in this application season compares your Scaled Score using test results from July 1, 2012 to June 30, 2017. The scaled scores of the new scoring change will only enter the Scaled Score-->Percentile Rank table for the 2019 application season. This means you might be most disadvantaged if you took the test in April 2017 but do not start applying until the Fall of 2018 (for 2019 start). However, this would only be 1 out of 5 years.
Note also that this "advantage" or "disadvantage" could only really occur if the distribution of scores for the Sept and Oct 2017 tests are skewed because half of the people did not know about the change. This might not even be the case. And the number of people who might gain such an advantage would be very small.
All in all, I do not expect even the worst case scenario to change someone's scaled score (and thus percentile rank) by very much, certainly not more than the large variance in the test scores.