Arguing that two humans' distinct interpretations of a single specification is stochastic is patently absurd and requires, I think, some shred of evidence.

I would argue that, in the absence of such evidence, we cannot conclude that the English-to-code conversion process is stochastic; it is merely not deterministic, which is altogether different.