[guided]At each stage $k = 1, 2, \ldots, n$, we perform a univariate division in the variable $X_k$. The input at stage $k$ is the remainder $r_{k-1}$ from the previous stage (with $r_0 = f$). We view $r_{k-1}$ as a polynomial in $X_k$ with coefficients in $S_k = R[X_1, \ldots, \widehat{X_k}, \ldots, X_n]$ and divide by $g_k(X_k)$, which is monic of degree $d_k$.
The critical observation is that dividing by $g_k$ in the variable $X_k$ does not disturb the partial degrees in any other variable $X_j$ with $j \neq k$. This is because $g_k \in R[X_k]$ involves only $X_k$: when we subtract a multiple $c(X_1, \ldots, \widehat{X_k}, \ldots, X_n) \cdot X_k^e \cdot g_k$ from the remainder, the coefficient $c$ comes from the existing coefficients of $r_{k-1}$ (which already satisfy $\deg_{X_j} c \leq \deg_{X_j} f$), and $g_k$ contributes no power of $X_j$. So the bound $\deg_{X_j} r_k \leq \deg_{X_j} f$ is maintained.
Similarly, previous divisions by $g_1, \ldots, g_{k-1}$ did not increase the partial degree in $X_k$: each $g_i$ for $i < k$ involves only $X_i$, not $X_k$. Therefore the bound $\deg_{X_k} r_{k-1} \leq \deg_{X_k} f$ holds at the start of stage $k$, and after division we get $\deg_{X_k} h_k \leq \deg_{X_k} f - d_k$.
For the total degree, since $r_{k-1}$ has total degree at most $\deg f$ and $g_k$ is monic of degree $d_k$, the quotient $h_k$ satisfies $\deg h_k \leq \deg f - d_k$ and the new remainder satisfies $\deg r_k \leq \deg f$.[/guided]