1: \begin{proof}

2: According to the structural assumptions, we may select $k >0$ sufficiently large (depending on $\delta$) so that, for each $\alpha=1,\ldots,\m$, the function $\varphi(y):= k - (1+|y|^2)^{1/2}$ is a smooth solution of

3: \begin{equation*}

4: \delta \varphi - \tr ( A_\alpha\!\left(y, \omega \right)\! D^2 \varphi ) + H_\alpha( p+ D \varphi,y,\omega)  \leq 0 \quad \mbox{in} \ \Rd.

5: \end{equation*}

6: Modify $u_\alpha$ by defining, for each $\ep > 0$,

7: \begin{equation*}

8: u_{\alpha,\ep}(y):= (1-\ep) u_\alpha(y) + \ep\varphi(y).

9: \end{equation*}

10: Formally, using the convexity of $H_\alpha$ and \eqref{fatrans}, for each $\alpha =1,\ldots,\m$, we have

11: \begin{equation*}

12: \delta u_{\alpha,\ep} - \tr ( A_\alpha\!\left(y, \omega \right)\! D^2u_{\alpha,\ep}  ) + H_\alpha( p+ Du_{\alpha,\ep} ,y,\omega) + f_\alpha\!\left( u_{1,\ep} , \ldots, u_{m,\ep} , \mu, y, \omega\right) \leq 0 \quad \mbox{in} \ \Rd.

13: \end{equation*}

14: This can be made rigorous  either by using the fact that $\varphi$ is smooth or by applying \cite[Lemma A.1]{AS}. Since $v_\alpha$ is bounded below and $u_{\alpha,\ep}(y) \rightarrow -\infty$ as $|y| \to \infty$, for all $R> 0$ sufficiently large and for each $\alpha=1,\ldots,\m$, we have

15: \begin{equation*}

16: u_{\alpha,\ep} \leq v_{\alpha} \quad \mbox{in} \ \Rd\setminus B_R.

17: \end{equation*}

18: It then follows from  \cite[Theorem 4.7]{IK} that, for each $\alpha=1,\ldots,\m$,

19: \begin{equation*}

20: u_{\alpha,\ep} \leq v_{\alpha} \quad \mbox{in} \ \Rd,

21: \end{equation*}

22: and, after sending $\ep \to 0$, the conclusion.

23: \end{proof}

24: