Do the reports / discussions relate to a software version (separate program loads) or the (re)configuration of a single ‘standard’ software?
A comparison with ‘older’ conventional aircraft, the production pre first flight testing could include installing specific testing devices such as control rod loading and angle measurements in the aircraft. These were not flight-worthy and thus had to be clearly identifiable and/or have ‘interference’ links preventing flight use. Similarly, failure conditions could be tested with system pin-outs or interference links (grnd/flt switches) which created abnormal situations which could not or would not be advisable to test in flight. The devices would again be clearly identified – big red flag, etc; which of course did not prevent the rare ‘surprise’ during first flights if some aspect was overlooked.
Are we to assume that similar activities with software loads/configurations are conducted during ground testing; i.e. not flight worthy programs / configurations. If so then the problem is perhaps more with software control/configuration opposed to the design/operation of the power-plant control system, but the latter should never be discounted.
I recall one ‘conventional’ design weakness which was not discovered during ground testing and resulted in all engines going sub idle at touchdown after first flight – ‘wt on wheels’/engine control logic, and a minor unrelated system failure. This could have happened in flight but fortunately the specific trigger circumstances were not encountered.