I'd like to know the lowest point $z_\min$ of an ellipse with radius $r_x, r_y$ in (Euclidian) XY that's tilted in XYZ - first rotated around X axis by $\gamma$, then rotated around Y axis by $\alpha$.
If it's only tilted around one axis, it's easy.
$$-z_\min = r_x|\sin(\alpha)|$$ (tx for pointing out Achille)
For the case $\alpha \neq 0$ and $\gamma \neq 0$, I thought about identifying the new direction and corresponding radius that lead to lowest point, but not sure exactly how to go about it.
What's the way to find $z_\min$ with both $\alpha \neq 0$ and $\gamma \neq 0$?
Instead of planar ellipse, let's us recall some results we know about ellipsoid.
Given any ellipsoid with semi-major axis $a,b,c$ and the corresponding symmetry axis pointing along direction $\hat{p},\hat{q},\hat{r}$ (represented as unit vectors). The equation of ellipsoid is given by
$$\frac{(\hat{p}\cdot\vec{x})^2}{a^2} + \frac{(\hat{q}\cdot\vec{x})^2}{b^2} + \frac{(\hat{r}\cdot\vec{x})^2}{c^2} = 1\tag{*1}$$
If $\hat{n}$ is any unit vector, the maximum extent of ellipsoid $(*1)$ along direction $\hat{n}$ is equal to
$$\max\big\{\; \hat{n}\cdot\vec{x} : \vec{x} \text{ satisfies } (*1) \;\big\} = \sqrt{ a^2 (\hat{p}\cdot\hat{n})^2 + b^2 (\hat{q}\cdot\hat{n})^2 + c^2 (\hat{r}\cdot\hat{n})^2 }\tag{*2}$$
If we treat the planar ellipse at hand as the limiting case of a flat ellipsoid and let $\hat{n}$ be the unit vector pointing towards the $-ve$ $z$-direction. We can read off the answer using $(*2)$ once we know the unit vectors $\hat{p}, \hat{q}, \hat{r}$ corresponds to the semi-major axis $a = r_x, b = r_y, c = 0$.
Start from same information as other answer, the transformation matrices for rotation about the $x$-axis for angle $\gamma$ and rotation about the $y$-axis for angle $\alpha$ are:
$$\mathcal{R}_{x,\gamma} = \begin{bmatrix} 1&0&0\\ 0&\cos(\gamma)&-\sin(\gamma)\\ 0&\sin(\gamma)&\cos(\gamma) \end{bmatrix} \quad\text{ and }\quad \mathcal{R}_{y,\alpha} = \begin{bmatrix} \cos(\alpha)&0&\sin(\alpha)\\ 0&1&0\\ -\sin(\alpha)&0&\cos(\alpha) \end{bmatrix}. $$ Multiply $\mathcal{R}_{x,\gamma}$ by $\mathcal{R}_{y,\alpha}$ from the left, the $3$ column vectors of the resulting matrix $$\mathcal{R}_{y,\alpha} \mathcal{R}_{x,\gamma} = \begin{bmatrix} \cos\alpha & \sin\alpha\sin\gamma & \sin\alpha\cos\gamma\\ 0 & \cos\gamma & -\sin\gamma\\ -\sin\alpha & \cos\alpha\sin\gamma & \cos\alpha\cos\gamma \end{bmatrix} $$ will be the $3$ unit vectors $\hat{p}$, $\hat{q}$, $\hat{r}$ we seek. i.e
$$ \hat{p} = \mathcal{R}_{y,\alpha} \mathcal{R}_{x,\gamma} \begin{bmatrix}1 \\ 0 \\ 0\end{bmatrix} = \begin{bmatrix} \cos\alpha \\ 0 \\ -\sin\alpha \end{bmatrix} \quad\text{ and }\quad \hat{q} = \mathcal{R}_{y,\alpha} \mathcal{R}_{x,\gamma} \begin{bmatrix}0 \\ 1 \\ 0\end{bmatrix} = \begin{bmatrix} \sin\alpha\sin\gamma\\ \cos\gamma\\ \cos\alpha\sin\gamma \end{bmatrix} $$
Throw this to $(*2)$ and recall $\hat{n} = (0,0,-1)$, we will obtain
$$ \begin{cases} \hat{p} \cdot \hat{n} &= \sin\alpha,\\ \hat{q} \cdot \hat{n} &= -\cos\alpha\sin\gamma \end{cases} \quad\implies\quad z_{min} = -\sqrt{ (r_x \sin\alpha)^2 + (r_y\cos\alpha\sin\gamma)^2} $$ Please note that if you tilt along one axis (say $\gamma = 0$) without any "shearing", then $z_{min} = -|r_x \sin\alpha|$ instead of $ -r_x\tan\alpha$.