I would guess that people leave because they realize that the amount of time and effort they have to put in is not worth it.
If this was true, we would expect that users would leave at higher rates when the rewards are lower. There have been times when the median post reward was substantially higher but users leave at the same rate (or even slightly higher rates).
Consistency of rewards could be an explanation which could be measured and tested for as a competing hypothesis.
You are assuming that all types of work are a static variable in this equation.
Blogging is a pretty niche activity. What percentage of the population should write for a living?
If your job is to pass butter... might not matter how much you're getting paid. Passing the butter is boring and not interesting. Tedious and monotonous even. Though the task at hand of passing the butter is easy, the psychological effort of continuing to grind away doing the same thing over and over with very little guarantees could have... sub-optimal results.
Just kidding I would totally expect more people to leave when the rewards were lower.
Very interesting analysis you have here.