In a recent answer, I found that this definite integral takes on an exact value of
$$\begin{align*} & \frac{4(a-1)}{\sqrt{(a-1)^2+1}} \left(a \sinh^{-1} \sqrt{a-1} - \sinh^{-1}(a-1)\right) \\ & = \frac{4(a-1)}{\sqrt{(a-1)^2+1}} \left(a \log\left(\sqrt{a-1}+\sqrt a\right) - \log\left(\sqrt{(a-1)^2+1} + a - 1\right)\right) \tag1 \end{align*}$$
for $a=\sqrt2=1+\tan\dfrac\pi8$. Numerical evidence {(1), (2)} suggests this is equivalent to the closed form given in the OP,
$$\begin{align*} & \log\left[\left(2a - 1 + 2\sqrt{2-a}\right)^\sqrt{2+a} \left(2a + 1 - 2\sqrt{2+a}\right)^\sqrt{2-a}\right] \\ &= \sqrt{a^2+a} \log\left(a^3-1+a^2\sqrt{a^2-a}\right) + \sqrt{a^2-a} \log\left(a^3+1-a^2\sqrt{a^2+a}\right) \tag2 \end{align*}$$
However, establishing the identity $(1)=(2)$ seems like a nontrivial task. It's not true for all real $a$ (plotted on the domain of $(2)$). Solving the underlying transcendental equation is hopeless as far as I can tell.
Question: How can one show that $(1)=(2)$ when $a=\sqrt2$?
(attempt/previous edit rolled up into self-answer)