Seeing as you already have reams of data at your fingertips, is it possible for you to do an analysis of the distribution of payments when you could only do 10 votes a day, compared to the period when the votes were increased to 40 a day?
I think the concensus is that the distribution worsened when it went to 40 a day - but it's hard to tell without data.
I will try to look into that some time this week.