I'd love to see a paper looking at this (e.g., how often is the "primary" author the last author or is the last author a better predictor of quality than the first, if you look back historically). I suspect this is very widespread issue.