I think this is just an argument over semantics, the "null hypothesis" is a perfectly valid model using the definition of the statistics community [0] (i.e. a collection of probability distributions over some sample space).
Approximately 65% of the worlds population suffers from a diminished ability to digest dairy after childhood, see http://ghr.nlm.nih.gov/condition/lactose-intolerance. That said lactose intolerance usually doesn't occur in people of European descent so I would imagine you happen to live in s special subgroup who has no such problem.
Not necessarily. Strictly mathematically a (non-negative) fraction q/p is just the equivalence class of all non-negative integer pairs (s, t), t != 0, such that
qt = sp
As such we may make q as small as possible by finding a representative in the equivalence class where the numerator is smallest (using the well ordering principle of the of the natural numbers). You can identify this pair uniquely (when the fraction is not 0) by using divisibility but that isn't required here.
[0] https://en.wikipedia.org/wiki/Statistical_model