The theory is that occasionally two fertilized cells (i.e. siblings or twins) merge together very early on to make one animal. Some of the parts of the body end up being from one of the siblings and some from the other.
In humans this didn't get noticed until they started doing maternity tests and finding moms that didn't match their babies. This is because they tested a sample from a part of the mom from one of the twins and her ovaries where from the other twin.
It might happen a lot in snakes but only get noticed once you start working with het morphs and have the potential to produce a chimera which mixes a morph and non morph sibling. In this theory the albino parts are from an albino sibling and the normal parts are from a het or possible het sibling (or likewise with axanthic).
Information that would help to determine how likely this is to be the correct explanation would be to know if these paradox animals always come out of breedings involving at least one het. If you could produce a paradox from albino X albino breeding then it would need to be something else.
It would also be interesting to know the breeding results from the paradox animals. Do paradox albinos sometimes produce offspring showing their reproductive organs to be het albino or perhaps normal or do they always breed like homozygous albinos?