Suppose $\mathcal{A} = \langle A, <, +, \cdot \rangle$ is a $\aleph_1$-saturated elementarily extension of the real field.
Is $SL(2, \mathbb{R})$ dense in $SL(2,A)$?
In compact groups one can find subgroups with infinitesimal entries, but is it still possible in $SL(2,A)$?
Turning my comment into an answer:
No, it's not. In fact, every "new" element of $SL(2,A)$ will have a neighborhood missing all of $SL(2,\mathbb{R})$ - and this uses merely the non-Archimedeanness of $A$.
Suppose $X\in M_2(A)\setminus M_2(\mathbb{R})$ (so $X$ has at least one non-real entry). Let $\epsilon$ be a "small enough" infinitesimal - specifically, let $\epsilon$ be smaller than (the absolute value of) the difference between each finite non-real entry $x$ of $X$ and the real part of $x$. The ball of radius $\epsilon$ around $X$ contains no elements of $M_2(\mathbb{R})$.
Now just make sure the determinant of $X$ is $1$. For example, fixing an infinitesimal $\delta$ we could use $$\begin{bmatrix}\delta & 0\\0 & {1\over\delta}\end{bmatrix}$$ (we could use $\epsilon=\delta^2$ as our corresponding radius).
EDIT: as Adayah observes, this in fact shows that $SL(2,\mathbb{R})$ is nowhere dense in $SL(2,A)$. As a further point of interest it's not hard to show that $SL(2,\mathbb{R})$ is closed as well (see their answer).