Specifically, are questions that specifically target the maximising of a specific character attribute value (at any cost or means) on-topic?
For example, for D&D3.5: "How to maximise a Wizard's effective Caster Level for the greatest number of spells, at character level 12?"
I think they would be on topic because they are specific and answerable, but also not because they can easily lead to one-upping answers?
I think ultimately they are on-topic—I certainly don’t think they cause any kind of problem that should lead to our deciding otherwise—but I would comment that Stack Exchange tends to not be a great format for performing optimization. The format works far better when an answer can point to and summarize the optimization work already done elsewhere. This is because an expert in the subject may easily be familiar with optimization work already done, and is in a good position to find and summarize it, but even a subject-matter expert here has no way of ensuring that a given approach is strictly optimal—as you say, there can always turn out to be another option to bump things up, or a better approach that hadn’t been considered. Optimization is often a collaborative, iterative process, and Stack Exchange doesn’t really do that very well.
Stack Exchange is, in this regard, far better at finding a good answer than it is at finding the best answer. That said, frequently a good answer is more useful—because finding the absolute best is hard. To block these questions on that basis would effectively be a matter of siding with the perfect against the good.
Given that there is a clear answer, it seems like these should be on-topic. One-upping is fine, because it will ultimately lead to the correct answer.
The question arises if there are multiple ways to do it and how to differentiate those that may devolve into opinion-based arguments.
