2) You can rarely use pure feed-forward in practice because, in order to keep the controlled variable at its setpoint, you would have to have perfect measurements of every variable that could affect the controlled variable (including, in some cases, effect of weather such as cold rain on an outside pipe or vessel). In a few rare cases when it is not necessary to hold the controlled variable at its setpoint and when feedback controlled variable is not measureable pure feed forward may be used. But this is rare.
1) I am not sure what you mean by "self destructive". But consider the case of a tank with flow in, which is measured, and flow out, which is controlled to be equal to the flow in. We start with the level at 50%. If the mesurement of the flow in was even slightly incorrect, eventually the time would come when the tank would become empty.
For this reason feedback is always used when feedforward is used. In most cases the result of the PID controller is added to a caculated feed forward signal.