Given a fitted polynomial using the numpy Polynomial convenience class (>1.14)

p = np.polynomial.Polynomial.fit(...)

what is the best way of saving and loading this fit?

There are three arguments which are needed to reproduce the object, namely p.coef, p.domain and p.window.
See this question for what happens when only the coefficients are loaded.

One could save all three as different files and re-load them, but that’s cumbersome and annoying.

My two initial ideas would be to save the entire object, either as a 0d-array

np.save("polyfit.npy", p, allow_pickle=True)
np.load("polyfit.npy", allow_pickle=True).item()

or as a 1d-array, as I honestly had to look up .item() for accessing the element.

np.save("polyfit.npy", [p], allow_pickle=True)
np.load("polyfit.npy", allow_pickle=True)[0]

Are there any recommended ways of doing this?
The documentation doesn’t seem to contain any good hint

https://numpy.org/doc/stable/reference/routines.polynomials.html
https://numpy.org/doc/stable/reference/routines.polynomials.classes.html