My theory is that analytics is to blame for all of this. You can track how many people clicked the button in your share header. You can't reliably track how many people left because of it (there could be a million reasons for leaving the page).

So your analytics shows an uptick in engagement by X%, and no clearly discernable downtick. So the header stays.

All it would need is an A/B test or two, but easier said than done...