My question is about this question and the users's answer to it.
Here's the statement of the compactness theorem: If $T$ is a first order theory in some language $L$. The $T$ has a model if and only if every finite subset of $T$ has a model.
One direction $\implies$ is trivial. The user proves the other direction as follows
Assume $T$ does not have a model. Then every sentence $\varphi$ in $L$ is provable from $T$. Let $\varphi$ be any sentence in $T$. Then there is a proof of $\lnot \varphi$ from $T$, $\varphi_1' , \dots, \varphi_n' = \lnot \varphi$. Now let $S = T \cap \{ \varphi, \varphi_1' , \dots, \varphi_n' \}$. Then $S$ is a subset of $T$ and $\varphi$ and $\lnot \varphi$ are provable from it. To see that $\lnot \varphi$ is provable from $S$, observe that $\varphi_i'$ used in the proof are each either a sentence in $T$ or a consequence of such or a formula that is tautologically true. If $S$ is empty, that is, none of them are in $T$, then $\lnot \varphi$ is provable without $T$ and the claim holds. If there are any $\varphi_i' \in S$ then all of them are axioms of $T$ so that by definition, $\lnot \varphi$ is provable from $S$.
My questions are
1.) Why is it clear that $\varphi$ is provable from $S$?
2.) If there are any $\varphi'_i \in S$, then why are all of them axioms of $T$?