Hi,
Looking to the safety aspect we may say ii would be good to preserve some functions of the A/P after anomalies are presented to inputs of the System.
AS not valid during the transitory (when AS is not briefly valid by any reason) is not necessary to "roll corrections".
My rationale is:
Every effort should be made from the designers to "present a machine" capable to help the crew
most of the time.
It seems the GIGO approach (used by IT professionals) could be better applied in an A/C for the sake of "Graceful Degradation" of the (effective) A/C, when PF is "suddenly inserted" in the "great loop".
The Stability of the System was affected by "tiny ice crystals" and IMHO the System could be designed (with about the same resources) to better "handle" similar situations.
Before, for example, better AS sensors are available.
(*) By "effective aircraft" or "effective system" we call the System + PF (when plane with it's automated resources has PF inserted in the "great loop"). Such occurred when PF started to hand fly F-GZCP that night.