I often think question that get closed in this way were asking for "too much".
Getting to know "why" something happens is important, we want to understand the world. The urge to ask for that is understandable. But if in the end we only get an opinion poll, we have missed our goal to build a knowledge repository. Every question must be posed so that everyone could answer it based on sufficient search of available information. We don't want answerers just to present their prior beliefs on the matter, we want to present solid information instead.
The pitfall to avoid here is clearly too much speculation. Some political analyses will include some speculation, let's say educated guesses, which might be interesting. But we must be careful to keep the level of speculation low in answers, otherwise Q&A here will simply be an as unreliable source as any other discussion board on the Internet.
My impression is that these questions could often be rescued if they were aimed differently (asking for less, the part that actually can be answered). Askers are of course free to try to make their closed questions ontopic and get them reopened.
For the example you give Is Putin trying to shore up his regime with a war? - we simply don't know what Putin's internal motivation is beyond his stated goals (and he might lie there), so we cannot answer it. But in this question there is the assumption that the reign of Putin in Russia might be in danger. Asking about that (something like "How secure is the position of Putin in Russia?") would be ontopic. And if you ask me that would also be more useful to know. I personally think of information as interesting and speculation as boring.
In some comments you suggest some kind of Bayesian inference to be able to make some wider statements about political issues. However some warnings there: Bayesian analysis requires a prior belief, so the answer would then unfortunately depend on who wrote it (everyone might give a different answer). But also the error would be huge if the underlying data is not good or not complete. And the data in politics is never really that well existing (there might be exceptions). Additionally statements in politics are frequently biased or wrong. A Bayesian analysis that by itself is reliable, would be extremely difficult, next to impossible.
Summary: Too much speculation doesn't result in good Q&A. Closed questions with this close reason should be reworded to ask for more factual issues (like the underlying assumptions). The interpretation of the available pieces of information should be left to everyone him/herself.