Just a quick question from an observation: is there a reason Tr[A] ~ d/E ?
Nevermind I found it in your comment on the math.SE post
I can show that expected trace is equal to $(1- \frac{1}{d})^d$ ...
So for large $d$, $(1- \frac{1}{d})^d \to \frac{1}{e}$.
