As noted in the comments, this algorithm will not terminate for certain choices of $a$. This is due to the fact that $R/I$ is not noetherian, which is equivalent to the existence of an infinite ascending chain of ideals.
It will terminate for other choices of $a \in R/I$ though, because being noetherian does not mean that if you start with any ideal there is an infinite ascending chain. For example if you take $a = 1$, the algorithm will clearly terminate.
However, given your infinite ascending chain you can find an $a \in R/I$ such that the algorithm never terminates. For example in your case if you pick any of the generators of those principal ideals $a = x_i$, the algorithm will never terminate because $x_i^{2^{i-1}} = x_{i+1}^{2^{i-1}}x_{i+1}^{2^{i-1}}$.
One way to reinterpret the algorithm is that if you consider the ideal generated by $a$, $(a) \subsetneq (b)$ because $a = a_1b$ with $a_1$ and $b$ non-units. This means $a \in (b)$ but $b \not\in (a)$ as if it were then $b = a_2 a$ for some element $a_2$ which is not a unit because $a_1$ is not. But then $a = a_1a_2 a$, which means $(a) = (a_1a_2a)$. This is only possible in your case if $a_1a_2$ is a unit, but this is not the case because $a_1$ and $a_2$ are not units. The algorithm will never terminate if we can always choose $b$ as above so that $(b) \subsetneq (c)$ for some other non-unit $c$. So having an $a$ such that this algorithm does not terminate implies the existence of an infinite ascending chain of principal ideals.
We noted above that an infinite ascending chain of principal ideals implies the non-termination of this algorithm for some $a$. So having an $a$ such that this algorithm does not terminate is equivalent to the existence of an infinite ascending chain of principal ideals (beginning with $(a)$).