Until recently, I wasn't aware of the subtle difference in $\to$ and $\implies$. I've read the previous answers, and want to put it all together to corroborate my understanding.
$P \to Q$ is a propositional statement formed by the binary logical connective $\to$ and two statements $P,Q.$ I see $P \to Q$ as a promise, thereby motivating its truth table where it's only false when $P$ is true and $Q$ is false.
Say, we are proving the theorem $$n\text{ is even}\implies n=2m\text{ for some integer }m.$$ Now, when we write at the end of our proof that $P \implies Q$, what we are saying is that if $P$ holds, then $Q$ also holds, if truth value of $P$ is 1, then truth value of $Q$ is also 1. That is, through our proof, we are denying there being any case of $P$ true and $Q$ false. So if I am to relate $P \implies Q$ to $P \to Q$, I'd say that $P \implies Q$ asserts that it cannot be the case that $P$ holds and $Q$ doesn't hold, thereby ultimately saying that $P \to Q$ is always true (including trivial vacuous cases), i.e. $P \to Q$ is a tautology.
$P \iff Q$ asserts if $P$ holds, then $Q$ also holds; and if $Q$ holds, then $P$ also holds. In relation to $P \leftrightarrow Q$, it says that $P \leftrightarrow Q$ is a tautology, i.e. $P \leftrightarrow Q$ is always true, thereby saying that the truth values of $P$ and $Q$ always go together as the same; and that it cannot be the case that truth values of $P$ and $Q$ are different. $P \iff Q$ is also called logical equivalence.
I've read that $\implies$ and $\iff$ are not connectives, but that they are making meta statements (statements about about propositions $P$ and $Q$). Are $\implies$ and $\iff$ just shorthand then for saying, "if you start with $P$, you can deduce $Q$" and "if you start with $Q$, you can deduce $P$" etc. ?
Yes.
No; based on your described context, $(P{\implies}Q)$ likely means that the conditional $(P \to Q)$ is universally true under the assumption of mathematical axioms/definitions, that is, that something like $$\forall x \;\forall y \;\Big(P(x,y) \to Q(x,y)\Big)\tag1$$ is mathematically true. Here, $(P \to Q)$ is not generally a tautology, and, typically, statement $(1)$ is not a logical validity.
In the context of your above sentence, the word always (equivalently: necessarily) is not well-scoped (and actually redundant), and tripping you up: in the current theory, $(P \to Q)$ is (always) true and $Q$ is (always) true whenever $P$ is true.
In short: mathematical truths are generally not tautologies.
Similarly, based on your described context, $(P{\iff}Q)$ asserts that $P$ and $Q$ are mathematically equivalent, but asserts neither that they are logically nor that they are tautologically equivalent.
Based on your described context, yes (assuming mathematical definitions and results).
Based on your described context, these two symbols are connectives but not material connectives.
The following answers that I wrote elaborate on this answer: