Documentation preview for this PR (built with commit 2d2592d; changes) is ready! 🎉
This preview will update shortly after each push to this PR.

As the added comment says

# The row space of w2 represents H^0(O(3D_0 - D1 - D2)), whose
# dimension is at least d0 - g + 1. Hence the codimension is at most
# 2*d0, and we cannot provide an expected_codim argument for mu_preimage.
w2 = self.mu_preimage(self.wV3, w1, self.mu_mat33)

we cannot guarantee the dimension of $H^0(O(3D_0 - D1 - D2))$ is exactly $d_0 - g +1$, according to the Riemann-Roch theorem, since the degree of the divisor can be as small as $g+1$ for the small model (the divisor is special). Hence $2d_0$ is not the expected codimension; $2d_0$ is only an upper bound for the codimension. By the way, expected_codim is used to speed up mu_preimage when the codimension is known.

Alright, looks good to me then.

I know that you know this because you reviewed #40221, but for anyone else reading, the long test warnings in the new doctest are fixed by #40221 which has been approved and is waiting to be merged into develop.

Thanks!