As a sideline view ... one of the main problems relates to the fact that certification stuff is idealised, not real world. The ICAO/regulatory task, progressively, is to reduce the delta.
(a) if the real world circumstances reasonably match the certification, then the outcome in the real world should be reasonably similar to what the (certification) book (AFM) suggests if the action sequence is done per the book.
(b) if not, then expect a variable and, perhaps, significantly variable, delta between the real world results and the certification suggestions.
How do both operator and commander handle this ? No simple answer but a lot of use is made of historical statistics and accident reports to come up with reasonably conservative SOPs for a start.
Winging it on the day is a recipe for disaster sooner or later. Overall the sensible use of historical statistics makes for a more reliable and predictable outcome. That's not to suggest that it will work ALL the time .. but, certainly, it will work MOST of the time.