The initial 9m/s estimate was years before actual M600 testing, and now real data exists (but is secret; as unfair business practice if Google is wrongly believed better in AWE). The hidden picture may be cut-in around 9m/s, and rated output now at rather higher wind velocity. Poor M600 low-wind performance may hide in simply motoring around the loop. The audio signal can reveal struggling in the upward phase.
Here is a more refined estimation heuristic-
Look up max power rating of the M600 motorgen, multiply by eight, then start subtracting parasitic factors (drag, mass penalty, electrical loss, overheating, etc). This may not give precise wind velocity, but does give a certain top power limit that is almost certainly being risked in testing, and may have caused the crash.
If M600 aerodymamic and motorgen estimates correspond, that’s good, but not very realistic. In real wind, that varies a lot moment to moment, the system may have high loss due to motoring through lulls and popping up to chase transient gusts. Practical estimation of M600 power capacity must account for borderline conditions.