DW (coloring by me)
(Perhaps more importantly) In Alternate Law (AF447), the computers defer to the input of the pilot - in this case the THS moved and stayed put because the PF was pretty much consistently pulling up. If he'd let go, then the "soft" protections would have started returning the THS towards the centre, and if the pilot had pushed forward for long enough, the THS would have moved forward significantly faster. In Direct/Abnormal Attitude Law (XL888), autotrim is off and all THS movement must be ordered by the pilot via the trim wheels.
Dozy, where do you have that (colored part) conclusion from? From year long discussions on several AF447 threads I remember, that the FBW system in ALT 2b, which was the mode AF 447 defaulted to, still strives for flight path stability in the pitch axis and thus in the absence of most all protections in the mentioned flight LAW would try to maintain the selected flightpath (which was most of the time an unhealthy climb) despite the unavailability of the required speed, although when the stick would have been released to neutral position. Remember, in normal LAW you change the flight path to a specific climb rate or bank angle, release the stick to neutral and the FBW logic would maintain this bank angle or climb rate until the pilot intervenes by a new command or the protections cut in. The same happens in ALT2b, except that there are no suitable protections active.
Your description would fit the pitch direct mode, (because it it would not have trimmed nose up anyway) which never latched.