Let $B$ be a Brownian motion and $s\wedge t := \min\{s,t\}$, $s\vee t := \max\{s,t\}$.
I stumbled over the following identity:
$$ \mathbb E[\exp(B(s\wedge t) + B(s\vee t))] \\=\mathbb E[\exp(B(s\wedge t) + B(s\wedge t) + [B(s\vee t) - B(s\wedge t)])] \\=\mathbb E[\exp(2\sqrt{(s\wedge t)} B(1))] \cdot \mathbb E[\exp(\sqrt{\vert t-s \vert}B(1))] $$
Evidently, this result can be obtained by using the scaling relation $B(t)=_d \sqrt{t}B(1)$, but if I try to apply this to the equation I get a different result.
How can one explain this identity?
$B(s \vee t)-B(s \wedge t)$ and $2B(s \wedge t)$ are indepedent since the Brownian motion has independent increments. Thus $$\mathbb{E}e(2B(s \wedge t) + B(s \vee t)-B(s \wedge t)) = \mathbb{E}e(2B(s \wedge t)) \cdot \mathbb{E}(B(s \vee t)-B(s \wedge t))$$
Moreover, obviously, $$B(s \wedge t) + B(s \wedge t) = 2 B(s \wedge t) \stackrel{d}{=} 2 \sqrt{s \wedge t} \cdot B(1)$$ by scaling property. Similarly, $$B(s \vee t)-B(s \wedge t) \stackrel{d}{=} B(s \vee t - s \wedge t) = B(|t-s|) \stackrel{d}{=} \sqrt{|t-s|} \cdot B(1)$$ where we used the stationarity of the increments (and again the scaling relation, for the last step).