Team A ships feature X and sets their KPI to some arbitrary measure of engagement. They miss, obviously, but instead of regrouping and hitting the drawing board, A doubles down and pressures Team B to point towards X in feature Y. A sees some marginal level of gain in engagement for X, obviously, so the intervention is deemed a success. 6mos later, Team A is asked to return the favor and add a modal pointing to new feature Z, per the request of Team B.
I don’t really know what the solution is except outside of careful org-wide watchdogging to ensure this sort of user-hostile engagement infighting gets nipped in the bud.