I do agree that (subjectively at least) there appear to be more closures of late. That said (and again subjectively), I haven't noticed a considerable increase in 'off-topic' closures. There have been more duplicates, and more 'unclear what you're asking' closures, and I think there have also been a few more 'too broad' closures as well.
I would also like to see some stats to support one position or another, but the simple fact is that most of the posts from the questioners on Meta, and most of the discussions in question comments appear to be around whether a question is distinct from another question (duplicate testing) or what the OP was really getting at (unclear questions). There has been increased frustration from newer members of our august body trying to ask questions that will get past the VTC process but relevancy has not been brought up anywhere near as much IMHO as these other two points.
The duplicate question problem is one that I expect to only increase. We have nearly 16k questions on the site at the time of writing this; that means that a lot of the 'good' questions have already been asked and writing a new good question is similar in scope (but not in scale) to writing a PhD thesis in a well researched area. How does one create something unique AND significant when all the 'framework' research is already complete? Creating a new good unique and significant question on this site is only going to get harder and there is a strong argument for direct editing by people with sufficient experience on the site to make these questions stand apart if the OP lacks the linguistic skills to do so. That same approach may also work for 'unclear' questions, provided intent or purpose can be established by someone competent at framing questions to assist the newcomer.
That said, I digress. The core question goes to relevancy of question and I think that the linked question was certainly a margin call. Yes, it's an engineering question, but it's also a 'missing ingredient' question.
How do I do X when Y (a critical ingredient of X) is missing?
In this case, the question is asking if there is a simple technology that's understandable by an ancient Aztec / Mayan style culture, that's workable with their current tools and artisanry, that can replicate the equivalent of an infrared door sensor.
Your other question about us being a catch-all is entirely relevant here, but so is the point that we have the capacity of being a catch-all because of the subjective nature of our topic rules. As such, we can't have the penny and the bun; either we accept that our rules are too subjective and tighten them up (removing a lot of potential margin call questions) or we leave it as is and accept that we have additional maintenance on the back end (so to speak) to clean up prospective questions, either through working with the OP to tighten them up or by direct closure. Essentially, we have to pick one approach.
Personally? I prefer the latter. Just because some of our newer folk have trouble articulating their question it doesn't follow that the question itself is poor. I'd rather have a pool of candidates that I filter through to find the gems than miss a good opportunity to increase our body of knowledge but I do completely understand that this is a high maintenance approach, and will not be favoured by all.
In so doing, we not only solve the on topic problem, but it also serves as a way to solve the unclear and duplicate question issues at the same time.