Possibly. But it also introduces a potential confounder.
> It's only futile if you conflate as "non-responders" the people who didn't respond on the first opportunity with the smaller number who wouldn't respond even given a second opportunity.
No, the non-responders are the ones who never respond, period. Which, as I said, means you can't sample them to see what they're like.
What you're doing, instead, is creating two different subsamples of responders: those who responded on the first try, and those who responded on the second try. Which might make things worse, since now you have an additional variable that you created by sending a second email to those who didn't respond the first time.
This is a better statement than any of mine of the point I've been trying to make.