Sounds fishy to me. The current sensors have about 50% QE. This means that one can only increase it twice, not 1,000 times.
In principle, "sensitivity" is not a well defined term when it comes to digital sensors. They are photon counting machines, missing every other photon, roughly speaking.
Yes, your thinking mirrors mine. Perhaps its a deeper electron well that holds more photons, but that implies a longer exposure. The description of "Wang said the key to his new sensor is the use of "light-trapping" nanostructures that use graphene as a base. The nanostructures hold onto light-generated electron particles for much longer than conventional sensors." is a bit vague and does not explain the predicted usefulness for consumer imaging.
I can see it useful for astronomy and night time imaging, perhaps even satellite imaging, but for a camera that is used to do high fps or video, I do not understand.
I agree with your statements - so I think it might be a sensor material which allows for 1000x the electrons per pixel increasing the dynamic range by 10 stops (8 stops for real products) - you will have perhaps a native sensitivity of ISO 400 - 8 stops lower sensitivity setting means ISO 4 (!!!) ...
For me it would be very interesting to have the freedom to choose very high DR or very long exposure times at high ambient lighting.
... perhaps "without flash" means that you can lift shadows due to the high DR without visible IQ loss? (this idea came up during writing ...)
The problem of durability could be solved by exchangeable sensors. Would be good idea just with CMOS sensors - I would like to convert my 20D to a B/W camera by removing the RGB Bayer pattern (and increase the sensitivity by a factor of two or three) but would like to have the chance to try the removal procedure on two or three sensors (not cameras).