>How did they control for couples that remove themselves from the sample set because they broke up due to bickering

Do they need to? Those are not couples anymore.

Besides, isn't it easy to track couples for several years, and see which humor each other, which bicker, and which break?

I don't know about you, but if I was with my partner " in a laboratory setting " I would certainly be on my best behavior.

Seriously, this is literally survivorship bias!

Only if non-survivors (broken up relationships) matter.

If the idea is "what happens while still in the context of a couple" then it doesn't. In fact that's already implicit in that the couple must "age".

That’s a good point. They did stay away from “the key to happy marriages is more humor” trope, so it really is just observational.

In effect, that couple stopped aging because that couple ceased to be.