I can't find it at the moment, but the last time I looked, attempted replications of an early "illegibility improves recall" study failed. (IIRC, that didn't use a custom font, but rather some de-contrasting/visual-noise applied.)

I see no cited papers here, just the claim, in the video, that "over 100 students" were used in testing to pick this font from several candidates.

I'm doubtful of any long-term value here. Even if "desirable difficulty" mechanism is real, I'd expect different readers to need wildly different levels of interference, and for their perceptions to adapt quickly to consistent letterforms. (So, to really get the benefit, you'd have an adjustable and dynamic level of perceptual-interference.)