If I've got it right, so far, the speculation is that, on a Cat3 autoland approach,the autopilot was inadvertently disengaged below re-engage height, then re-engaged, the ILS deviation wasn't monitored, the lack of correct annunciators was ignored, the correct references weren't available at decision and a go around wasn't initiated in sufficient time to avoid heavy ground contact.
I find that sequence too difficult to believe of two professional pilots.