Back to old obsession:- what exactly was happening just before the incident started - were the controls acting on bad data? I think its possible but only for a few seconds before the air data correctly flagged as inconsistent.
One possible failure mode of pitot tubes when encountering ice is that the drain holes block first and cause a small over-read, then under-read as the total port blocks. With the air speed under auto-control this might not be seen on the Mach No - but may be seen to be calculated as change in wind speed.
Is it it just a coincidence that it all seems to be coincident with the selected Mach drop from M0.82 to M0.80? This seems to have been commanded at 2:09:50, unfortunately the plot of "Navigation Parameters" on English P109 is corrupt while the selected Mach is 0.82 - as mentioned before it appears they are plotting 0 when no update.
The engine N1s start to ramp down from 2:10:00, but seem to have little effect on the indicated speed - but presumably is responsible for the co-ordinated drop in pitch angle? The longitudinal acceleration suggests the plane is slowing (although I don't understand the negative bias on the plot).
It appears the ISIS calibrated airspeed when recovered after the dip is "stuck false high" in period 2:10:10 to 2:10:13 - during which the engine N1s are locked low - but groundspeed doesn't drop off much (perhaps due to plot scale).
It's not all consistent to my eyes, and I hope the BEA can recover more detail data the control hardware around this time (especially the right hand display indications) - BUT - I think you need significant shifts in wind speed/temperature to make it all add up - what if they "nearly missed" a Cumulonimbus?