>>hoseja+(OP)
I think that's widely understood to be 'sort of true'. Not necessarily a literal knot, but a geometric property - like what I mentioned, where integrating around sees a net divergence/curl/etc in any reference frame. That's the reason that QM doesn't predict exact locations for particles: the discontinuity doesn't have a location, it just exists in a region.
(I don't exactly understand if this 'is' a knot, in a sense. I guess it is.)