Similarly it may be the case that happy men end up in marriages rather than men end up in marriages and become happy as a result.

You're still stuck in the trap. It isn't either of those. It's that the fail state is defined out of the equation: Survivorship bias is built into the question.

"We polled 200 living survivors of gun violence, and found that none of them died. Therefore, we can assume it's likely that gun violence doesn't cause deaths."

That words it a bit more hyperbolically, but also a bit more clearly for you. The men who are ruined are hidden away in the divorce, separated, widower, or suicide statistics, not the happily married in a functional marriage statistic, so it doesn't matter how many happily married in a functional marriage men you poll, you're not going to find data on the ones who hit that fail state. [EDIT: And when 50% of marriages end in divorce, letting that 50% of the pollable data in question get allocated to "single" men instead, and then summarizing that these divorced, broke, broken, abused "single" men are clearly miserable, is the worst lying with statistics.]