Thanks KuRi & 2020_SRS_Commuter for you answer, I'll study all these points.
The Rear wheel is correctly aligned, no problem with that.
My problem is exactly the same as this one :
https://www.electricmotorcycleforum.com/boards/index.php?topic=12714.0By testing the frequency of the belt with different positions of the wheel with the motorcycle on the lateral crutch, I find between 58 and 68 Hz. So now I have just understood that the cause is the crown which cannot be perfectly circular and have its center perfectly in the axis of rotation of the wheel.
The old belt (which has a small tear) has been mounted again yesterday, the sound of resonance is much less, there is therefore clearly a frequency of resonance specific to each belt, which must come to resonate with the couple gable / crown and their respective shapes and centering.
The average tension adjustment of the belt is secondary for this problem, it plays, but not a lot (to tighten the belt aggravates this resonance problem a little too), while the difference between 2 different belts therefore plays a lot on the phenomenon.
I was able to have a belt to go and an another belt on the return, so the comparison was easy: the set of belt noises is very different from one to the other at different speeds / acceleration / decelerations => It is obvious, it is not embarrassing apart so this big resonance and 1st acceleration in the maximum, rolling it then does not come, except at the start of the climb so, so it comes back to a strong couple required.
There will be other handles by the dealership to try to resolve this resonance with the new belt. I will post at that time.