Within physics you have astrophysics or cosmology which are also historical.

That would be a reasonable statement.

And, to a certain extent, stationarity and ergodicity are model-dependent features.

As a rule, stationarity and ergodicity are directly observable on experimental data if you have it, and vice versa. It's by no means a given in all theoretical models, but it's a real observable phenomenon regardless of subjective prior assumptions.

It's not that "postmodern" "textual analysis" doesn't apply to physics, it is that when studying physics one has to take into account the repeatability.

I never claimed it couldn't be applied to physics, rather I fail to see its value when an oracle exists which spits out facts for any well chosen question. Thus physics is not limited by the insights of postmodernism.

So "repeatability" is a special quality that applies to a subset of human knowledge.

Another good word is "interactivity". In an experimental science, we can choose the questions we want to ask, and receive answers from the world. In a "historical" science, we must accept the answers we are given, with little or no choice in the questions. In physics, much effort is spent designing experiments to isolate the bits we care about, in archaeology we cannot ask what the ancient Greeks would have been like if they had had television.

--
$E(X_t|F_s) = X_s,\quad t > s$