At the very minimum, this reduces the chance of fraud to 0%. Clearly it is possible to replicate something similar to the levitation shown in the original video.
Even if it isn't fraudulent it still may be false, but it seems quite certain that the authors believe in their own claim.
I'd also abstain from differentiating "fraud" and "error" as this is irrelevant to whether the paper is reproducible.
What we care about is true positives and negatives. But we'll also get false positives and negatives.
So replication will have to happen at scale. Individual outcomes don't matter much.