The RAAS technical description in Tech Log raises further questions and provides a interesting scenario which might relate to this accident.
The RAAS 'Long Landing" alert is only given in the air between 100ft and 5ft RA.
The distance down the runway where the alert is given is pre-selectable by the operator according to aircraft type and runways used.
The position information comes from GPS (GPS/FMS or internal GPS in EGPWS? Which is more accurate?). The warning distance appears to be based on GPS, but could involve ground speed-distance calculation from either the threshold or runway end.
The 'standard' distance is 2000ft or 25% of the runway. 2000ft might be a bit short for a 777, particularly when regularly operating in tailwinds: 2000ft is 25% of an 8000ft runway, thus some compromise is required. The distance chosen by this operator is not known, nor the option of fixed distance or %.
The 'only in the air' aspect is very interesting.
Requiring a go-around after an alert would, with simple assumption, always be an 'airborne' GA where TOGA will provide auto thrust commands, even if the aircraft subsequently touches the runway as the AT will remain engaged.
However, if the alert is very close to the ground, above 5ft, it would be possible - likely, that the TOGA selection could be made below 2ft, where the auto thrust (TOGA) is inhibited.
More frequent alerts might be expected in this situation because the aircraft is always reducing the distance to the long landing point during a normal landing.
In such a scenario a crew can follow the procedures, expect normal auto thrust operation based on the (incomplete) Boeing/FCOM system description - "TOGA is inhibited on ground". The crews expectation may be enhanced by normal operations (always use AT).
Although the manual thrust arguments are well founded, it is probable that a 777 PF never manually advances the thrust levers - not during takeoff, and nor during a GA (do operators encourage two handed finesse for pitch control to avoid a tail strike?).
During takeoff the crew focus is on power and speed, the task is to fly.
During a GA the task has to be quickly reformulated from the landing objective; PF has to fly, PM select flaps, check FMS, FMA, and hearing, noting, and responding to ATC clearance ... and verifying that the thrust is set.
A 10-12 sec delay in identifying the unexpected AT operation might be reasonable given the workload and behavioual influences.
And as much as the manufacturer, regulator, management, (or some Ppruners) might expect the crew to recall the FCOM notes on page 4.29.17. line 12, or the textural rejected landing procedure, humans, pilot's do not function that way, particularly in very rare and surprising situations, where procedures have been actioned correctly, ... with a minor oversight which subsequently appears to have been identified, but a little too late.
This is an 'as designed' situation, with opportunity for misleading system operation, and probably with the gross assumption that the crew will manage error avoidance, detection, and mitigation: in a time scale much shorter than it takes to read this post.
Pilots have a responsibility for safety, but just because they are the last line of defence, their responsibility does not include the responsibilities of those who have much more time and influence to consider the situation; the regulators, manufacturers, and operator.