I think the point here is that there are Two issues mixed up here; training and qualification.
By definition training involves the possibility of the student failing at first attempts to perform the required action. Hence there must be an allowance made for the instructor to rectify mistakes in good time.
As for qualification. I fail to understand the logic in assuming that if a pilot can handle EFATO once during a test, they can therefore handle the same failure, possibly in a different environment in future.
Then there is the argument for professionals about the risk mitigation possibilities in actually performing an EFATO. Clearly, training fatalities suggest that a realistic practice is as dangerous as the real thing.
To put that another way, does one test the operation of a firearm safety catch by loading the weapon, pointing it at your head and attempting to pull the trigger?