I have a question about techniques used in determining the stratification over which a constructible sheaf falls into even constant pieces demonstrated on this example from Wikipedia.
Let $f:X = \text{Proj}(\frac{\mathbb{C}[s,t][x,y,z]}{(s \cdot t \cdot f(x,y,z))}) $ $ \to \text{Spec}(\mathbb{C}[s,t])$
given canonically, where $f(x,y,z)$ homogeneous polynomial determining a smooth plane curve $C_f \subset \mathbb{P}^2$. Then the fibers over closed subscheme $V(s \cdot t)$ are isomorphic to $\mathbb{P}^2$, and over the the complement $\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)$ are naturally isomorphic to $C_f$.
Consider the derived pushforward $\mathbf{R}f_!(\underline{\mathbb{Q}}_X)$ of the constant sheaf $\underline{\mathbb{Q}}_X$.
Then it is claimed that for restrictions over $V(s \cdot t)$ holds
$\mathbf{R}^kf_!(\underline{\mathbb{Q}}_X)|_{\mathbb{V}(st)} = \underline{\mathbb{Q}}_{\mathbb{V}(st)} \ \ \ ,k=0,2,4 \\ \mathbf{R}^kf_!(\underline{\mathbb{Q}}_X)|_{\mathbb{V}(st)} = \underline{0}_{\mathbb{V}(st)} \ \ \ ,else $
and over complement $\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)$
$\mathbf{R}^kf_!(\underline{\mathbb{Q}}_X)|_{\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)} = \underline{\mathbb{Q}}_{\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)} \ \ \ ,k=0,2 \\ \mathbf{R}^kf_!(\underline{\mathbb{Q}}_X)|_{\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)} = \underline{\mathbb{Q}}^{\oplus 2g}_{\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)} \ \ \ ,k=1 \\ \mathbf{R}^kf_!(\underline{\mathbb{Q}}_X)|_{\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)} = \underline{0}_{\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)} \ \ \ ,else $
where $g$ the genus of $C_f$.
Question: Could it be elaborated how to obtain this result and which techniques are involved?
Since $f$ is proper, the fibers are easy to determine in light of proper base change: $(\mathbf{R}^kf_!(\underline{\mathbb{Q}}_X))_y= H^i(X_y, \underline{\mathbb{Q}}_{X_y})$.
But even if this implies that on $\text{Spec}(\mathbb{C}[s,t])- V(s \cdot t)$ and $V(s \cdot t)$ the fibers not change, this a priori not implies that restricted to these two pieces the pushforward sheaves are constant there. Naively, they could be also only locally constant there.
Why is this here not the case, and the stratification by $V(s \cdot t)$ decomposes the $\mathbf{R}^kf_!(\underline{\mathbb{Q}}_X)$ already into constant sheaves there?