I have attempted this problem and came up with the following method to solve it
One can show that the sequence $z_{n+1}=\sqrt{a+z_{n}}$ is monotone I assumed it's increasing as a case study , therefore relying on the monotone convergence theorem
I tried to estimate a possible candidate for a bound, assuming $(z_{n})$ converges we get
$$\lim_{n}z_{n+1}=\lim_{n}\sqrt{a+z_{n}}\implies z=\sqrt{a+z}$$
therefore
$$z^2-z-a=0\implies z=\frac{1\pm\sqrt{1+4a}}{2}=\frac{1}{2}\pm \sqrt{\frac{1}{4}+a}$$
If the sequence converges choose $z=\frac{1}{2}+\sqrt{1/4+a}$ this would be the sequences supremum and $\frac{1}{2}\pm \sqrt{1/4+a}\leq 4+a$
therefore I made the claim that $z_{n}\leq a+4, \; \forall n\in\mathbb{N}$, and I tried to prove it as follows
assume there exists $n_{0}\in\mathbb{N}$ such that $z_{n_{0}}>a+4$ then
$$S=\{n\in\mathbb{N}:z_{n}>a+4\}\neq \emptyset$$
thus by the well ordering principle it attains a least element say $m$ but then
$$\begin{align}z_{m}&>a+4 \\ \sqrt{a+z_{m-1}}&>a+4 \\ a+z_{m-1}&>a^2+8a+16 \\ z_{m-1}>(a+4)+(a^{2}+6a+12)&>a+4 \end{align}$$
This would contradict the choose of $m$ and create an infinite descent therefore, $(z_{n})$ is a monotone and bounded sequence thus it's convergent and it's limit is the supremum suggested above