The system proposed in this piece could probably be safely implemented, but it would much more challenging to do than the author seems to indicate. There are a variety of failure modes that they ignore.
The author also glosses over a 20% power loss (to heating), which would cost money and bulk to dissipate.
I particularly dislike how this article tries to sweep the power loss issue under the rug: The first mention of loss says that the isolation link is "responsible for about 50 percent of the charger’s power loss". The second mention of loss says "the power loss would be less than 20 percent". They want you to think that those numbers are comparable, and that they've reduced loss from 50% to 20%.
The article needs absolute efficiency figures to know if this is a good idea. I also wonder if better conversation electronics could make a difference. I'm guessing there is some tradeoff between cost of conversion and volume/weight/efficiency.
The other question is how this fits in with regenerative braking. Is this power conversion circuit dual purpose?
The author also glosses over a 20% power loss (to heating), which would cost money and bulk to dissipate.