The official infection and death numbers in China appear to be completely fabricated as they followed an almost perfect quadratic progression: https://old.reddit.com/r/dataisbeautiful/comments/ez13dv/oc_... None of the various measures taken to contain the outbreak have affected those numbers. This apparently has happened before with organ donation data: https://bmcmedethics.biomedcentral.com/articles/10.1186/s129...
Let's take a look at that paper you linked. The supposedly damning point is that you can fit the data extremely well with a quadratic, in the sense that R^2 is high. However, the data points are rapidly increasing, and because of the way R^2 works, that means only the last ~4 data points have any impact on R^2 at all. The statement of the paper boils down to the claim that you can fit 4 data points on a smooth curve accurately using a 3-parameter fit whose form you got to choose. This is obviously true, for any 4 data points.
It's too bad that reason goes out the window whenever China is involved.