I think it should be possible to prove, but since I can not for the life of me find a natural deduction proof that prove it, I have concluded that it should be possible to find a model that prove it is not valid.
The problem is that if I for example have a model $M$ where $A = \{1,2\}$, $S^M = \{1\}$ and $T^M = \{2\}$ so that $M \vDash \forall x (S(x) \lor T(x))$ is valid, then also $M \vDash \forall x S(x) \lor \forall x T(x)$ is valid. Can not get around it.
Am I correct that it is valid? If so how should I approach it in order to construct a proof in natural deduction?
For example, let $S(x)$ iff $x$ is sad. and $T(x)$ iff $x$ is tall.
And we know that
Everyone is sad or tall
$$∀x(S(x)∨T(x))$$
Does this implies
Everyone is sad or everyone is tall ? i.e.
$$∀xS(x)∨∀xT(x)$$
What about in a world with two people, one of them is tall but not sad, another is sad but not tall? In this case $∀x(S(x)∨T(x))$ hold, but $∀xS(x)∨∀xT(x)$ does not hold, this give a counter example.
This example is actually equivalent to your model, $$A=\{1,2\} , S^M=\{1\}, T^M=\{2\}$$ There are two people in world $A$, i.e. '$1$' is sad but not tall, '$2$' is tall but not sad. So it's not the case that Everyone is sad or everyone is tall, that $∀xS(x)∨∀xT(x)$ does not hold.