For example, let $$p(t) = (t-1)(t-2) = t^2-3t+2$$ be the monomial in question, with roots $x_1 = 1$ and $x_2 = 2$.
The elementary symmetric polynomials in terms of these roots are: $$x_1+x_2 = 3$$ and $$x_1 x_2 = 2$$
These polynomials are equal under two permutations: identity and transposition, just like the case for distinct irrational roots: $$x_2+x_1 = 3$$ and $$x_2 x_1 = 2$$
What does the approach in these sources use to justify, then, that the permutation group of the roots of $p$ is trivial?
I understand why this is the case when the problem is generalized to automorphisms on field extensions, where the splitting field over the rationals of a polynomial with rational roots is still the rationals, and hence only admits the trivial field automorphism. However, how does this translate to the classical case where permutations of symmetric polynomials in terms of the roots are considered instead?
Starting from a formulation using field automorphisms, I understand why the Galois group is trivial.
As far as I'm aware, however, the formulation in terms of symmetric polynomials of the roots came much earlier; so the question is not why is the Galois group trivial for this case, to which the answer using a modern approach is clear to me, but rather what assumptions are being made in the sources, that I am missing, that justify the only allowed permutations being trivial in the older symmetric polynomial approach?