Davies dealt briefly with the notion of perfection as well, in offering the thought that air transports must be realistically designed for average skill.
While the goal of perfection is entirely appropriate for the military due mission demands, commercial air transport, particularly in non-state-supported corporations, have neither (relatively) unlimited resources nor does the population base (overall, and of those with pilot licences), support a "top-gun" standard for a commercial enterprise.
As we are increasingly aware, the very opposite is now in place where just enough skill, completed by OTJ-training and MCPL licencing are the twin standards for commercial air transportation. Even that system is not supplying sufficient pilots for rapid growth we are seeing in some parts of the world. Nor is "automation" in any way a supplement for such shortages, particularly if one does not thoroughly know one's aircraft.
That said, alf5071h has, I believe, a good point that manual flying skills need not be ready or practised for rare, complex manoeuvres but should be ready for bread-and-butter operations that at least stabilize the aircraft.
Long-discussed and still relevant, the PF on AF447 certainly had that level of skill as he very quickly and successfully got the roll oscillations under control. Illustrative and in the present context of the thread, the problem was cognitive, not an inability to fly manually. The problem was not comprehending the situation the aircraft was in, which was a non-event and certainly not an emergency, as aircraft which lose airspeed indications remain stable in cruise so long as pitch and power settings do not change.