I don't necessarily agree that zeta is the most relevant stastic to measure rewards, but i think it would be interesting to see a real statistical analysis of late july zeta. vs zeta right now.
A more valid stastical metric for the success of reward distribution between mid july and now would be weekly lists of users, grouped according to standard deviation from the mean for rewards, and how much of the reward pool went to which standard deviations.
I suspect (though i havent yet been able to figure out how to pull out the information in an automated way) that neither metric would support your position.