First, we will show by induction on n that \mu_i(x_n)>0 for all 0\leq i\leq m-1 and all n\geq0. It is clear that \mu_i(x_j)=1 for 0\leq i,j \leq m-1. Now, we fix i\in \{0,\ldots,m-1\} and n\geq0. Assume that we have \mu_i(x_j)>0 for all j\leq n+m-1. We will show that \mu_i(x_{n+m})>0. By equation (?), we have
\mu_i(x_{n+m}x_n)=\mu_i\left( \sum_{s=1}^{m-1}\sum_{t=1}^{m-1}c_{s,t}x_{n+s}x_{n+t} \right).
Now, we look at the x_i-adic valuation of the sum
S=\sum_{s=1}^{m-1}\sum_{t=1}^{m-1}c_{s,t}x_{n+s}x_{n+t}.
Let
\nu=\min\{v_{x_i}(x_{n+s}x_{n+t})\mid 1\leq s,t \leq m-1, c_{s,t}\neq0 \}
be the minimal valuation among the terms of the sum S. We split the sum S into two summands S=S_1+S_2, where S_1 is the sum of the terms c_{s,t}x_{n+s}x_{n+t} of valuation \nu, and S_2 is the sum of the terms of higher valuation. We claim that v_{x_i}(S_1)=\nu; indeed, passing to the residue field of the valuation v_{x_i}, and looking at the x_i-adic images of the terms in S_1, we know by induction that the leading terms \mu_i(c_{s,t}x_{n+s}x_{n+t}) are strictly positive. It follows that the leading term of the sum of these x_i-adic images is also positive, and in particular does not vanish. Hence indeed, v_{x_i}(S_1)=\nu and \mu_i(S_1)>0. Since the terms in S_2 have higher valuation, we have v_{x_i}(S)=v_{x_i}(S_1) and \mu_i(S)=\mu_i(S_1)>0. Hence,
\mu_i(x_{n+m})\mu_i(x_n)=\mu_i(x_{n+m}x_n)>0.
Therefore, \mu_i(x_{n+m})>0. Furthermore, we obtain that
\begin{aligned}
v_{x_i}(x_{n+m})=\min\{v_{x_i}(x_{n+s})+v_{x_i}(x_{n+t}) \mid c_{s,t}\neq 0 \}-v_{x_i}(x_n),
\end{aligned}
for 1\leq s,t\leq m-1.
Observe that for the initial variables we have x_i-adic valuations for 0\leq i,j \leq m-1
v_{x_i}(x_j)=\begin{cases}
0 &\text{ if } j\neq i,\\
1 &\text{ otherwise}.
\end{cases}
We claim that for 0\le i\le m-1
v_{x_i}(x_n)=
\begin{cases}
0 & \text{if } 0\le n\le m+i-1,\ n\neq i,\\
1 & \text{if } n=i,\\
-1 & \text{if } n=m+i.
\end{cases}
Indeed, fix 0\leq i \leq m-1, and let n=n'+m with 0\leq n'\leq i. Note that P is not divisible by x_{i-n'}. Thus, there is a pair (s,t)\in C with i-n'\notin \{s,t\}. Therefore, by induction on n' we get that the minimum valuation in (?) is 0. Hence, v_{x_i}(x_{n'+m})=-v_{x_i}(x_{n'}) for 0\leq n'\leq i which proves the claim.
Now set \kappa_n^{(i)}:=-v_{x_i}(x_n) for n\neq i and \kappa_i^{(i)}:=0. Then
\kappa_0^{(i)}=\cdots=\kappa_{m-1+i}^{(i)}=0,\qquad \kappa_{m+i}^{(i)}=1,
and, using (?) and \kappa_n^{(i)}:=-v_{x_i}(x_n) we obtain
\kappa^{(i)}_{n+m} = \max_{(s,t)\in C}\big(\kappa^{(i)}_{n+s}+\kappa^{(i)}_{n+t}\big) - \kappa^{(i)}_{n}
for n\geq i+1.
Now, we will observe that \kappa_n^{(i)}\geq 0 for all n\geq 0. By assumption, (t,m-t)\in C, hence
\kappa_{n+m}^{(i)}+\kappa_n^{(i)}\geq \kappa_{n+t}^{(i)}+\kappa_{n+m-t}^{(i)}, \qquad n\geq i+1.
We show by strong induction on N that \kappa_N^{(i)}\geq \kappa_{N-t}^{(i)} for N\geq t and \kappa_N^{(i)}\geq \kappa_{N-(m-t)}^{(i)} for N\geq m-t. If N\leq m+i, this is clear from the initial values, since the only nonzero value among \kappa_0^{(i)},\ldots,\kappa_{m+i}^{(i)} is \kappa_{m+i}^{(i)}=1. If N\geq m+i+1, write N=n+m.
Now assume that for n'< N we have
\kappa^{(i)}_{n'}\geq \kappa^{(i)}_{n'-t}, \text { for } n'\geq t \text{ and } \kappa^{(i)}_{n'}\geq \kappa^{(i)}_{n'-(m-t)} \text{ for } n'\geq m-t.
Thus, we get
\kappa_{N}^{(i)}-\kappa_{N-t}^{(i)}\geq \kappa_{n+t}^{(i)}-\kappa_n^{(i)}\geq 0
\quad\text{and}\quad
\kappa_{N}^{(i)}-\kappa_{N-(m-t)}^{(i)}\geq \kappa_{n+m-t}^{(i)}-\kappa_n^{(i)}\geq 0,
since n+t<N and n+m-t<N. Now, for N\geq 0 write N=qt+r with 0\leq r<t. Then
\kappa_N^{(i)}\geq\kappa_{N-t}^{(i)}\geq\cdots\geq\kappa_r^{(i)}=0.
Consequently, v_{x_i}(x_n)\leq 0 for all n\neq i since \kappa_n^{(i)}=-v_{x_i}(x_n) for n \neq i.
Note that \{x_n\}_{n\geq0} has the Laurent property. Therefore, we can write x_n uniquely in the form
x_n=\dfrac{g_n}{x_0^{k_n^{(0)}}\cdots x_{m-1}^{k_n^{(m-1)}}}, \qquad n\geq 0,
where g_n is a polynomial in x_0,\ldots,x_{m-1}, the k_{n}^{(i)} are non-negative integers, and for each 0\leq i\leq m-1 either k_n^{(i)}=0 or x_i\nmid g_n. In particular,
v_{x_i}(x_n)=v_{x_i}(g_n)-k_n^{(i)}, \qquad v_{x_i}(g_n)\geq 0.
Now let n\neq i. If k_n^{(i)}>0, then x_i\nmid g_n, so v_{x_i}(g_n)=0 and k_n^{(i)}=-v_{x_i}(x_n). If k_n^{(i)}=0, then v_{x_i}(g_n)=v_{x_i}(x_n)\leq 0. Thus v_{x_i}(g_n)=0, which means x_i\nmid g_n. Therefore, k_n^{(i)}=0=-v_{x_i}(x_n). Thus, for n\neq i we have x_i\nmid g_n and
k_n^{(i)}=-v_{x_i}(x_n)=\kappa_n^{(i)}.
For n=i we have x_i=x_i/1, so k_i^{(i)}=0=\kappa_i^{(i)}. Therefore k_n^{(i)}=\kappa_n^{(i)} for all n\geq 0, and the sequences \{k_n^{(i)}\}_{n\geq0} satisfy the initial conditions and the recurrence in the statement.
Finally, we can compute the degree of iterates of the map f. We write the n^{\text{th}} iterate of the map f as
f^n\colon (x_0,\ldots,x_{m-1})\mapsto (x_n,\ldots,x_{n+m-1}).
We denote m_n^{(i)}:=\max\{k_j^{(i)}\mid n\leq j\leq n+m-1 \} for 0\leq i \leq m-1. In projective coordinates, the map f^n takes the form
\begin{aligned}
(x_0:\dots:x_{m-1}:y)\mapsto \bigg(&G_nx_0^{(m_n^{(0)}-k_{n}^{(0)})}\cdots x_{m-1}^{(m_{n}^{(m-1)}-k_{n}^{(m-1)})}:\\
&\dots:\\
&G_{n+m-1}x_0^{(m_n^{(0)}-k_{n+m-1}^{(0)})}\cdots x_{m-1}^{(m_{n}^{(m-1)}-k_{n+m-1}^{(m-1)})}:\\
&x_0^{m_n^{(0)}}\cdots x_{m-1}^{m_n^{(m-1)}}y \bigg),
\end{aligned}
where G_j=y^{\deg(g_j)}g_j\left(\frac{x_0}{y},\dots,\frac{x_{m-1}}{y}\right) are homogenizations of the polynomials g_j’s, and are of degree
k_j^{(0)}+\dots+k_j^{(m-1)}+1.
We claim that the projective coordinates of f^n have no common factor for all n\geq 0. Indeed, y can not be a common factor since G_j’s are homogenizations of the polynomials g_j’s.
Now, if m_n^{(i)}>0 for any 0\leq i\leq m-1, then m_n^{(i)}=k_j^{(i)}>0 for some n\leq j \leq n+m-1. Thus, x_i \nmid g_j, so x_i \nmid G_j. Now we look at j^\text{th} projective coordinate of the f^n and we see that
x_i \nmid G_jx_0^{m_n^{(0)}-k_j^{(0)}}\dots x_{i-1}^{m_n^{(i-1)}-k_j^{(i-1)}}x_{i+1}^{m_n^{(i+1)}-k_j^{(i+1)}}\dots x_{m-1}^{m_n^{(m-1)}-k_j^{(m-1)}},
since m_n^{(i)}=k_j^{(i)}. Additionally, if m_n^{(i)}=0, then x_i \nmid x_0^{m_n^{(0)}}\cdots x_{m-1}^{m_n^{(m-1)}}y. Hence, x_i can not be a common factor for projective coordinates. Therefore,
\deg(f^n)=m_n^{(0)}+\cdots+m_n^{(m-1)}+1.