Really interesting counter-question you posed. It appears to then be a decision about Path's conviction to how things SHOULD be designed vs. immediate conversion lifts (and how significant those lifts were/what future engagement was like/other conversion and engagement metrics).
For example, I imagine Facebook has tested designs that increased user engagement but didn't fit with a bigger picture strategy or violated some core tenets. They probably scrapped it and stuck to their conviction in a lot of cases when there were alluring potential upticks in conversions with alternate designs.