You didn't list any particular questions as examples (the one question you link to never received a close vote), so I'm going to list some recent "too localized"-closed questions and discuss them. Here is a list of every question closed as "too localized" during the last 48 hours or so (as of when I posted this answer):
That's six questions closed as "too localized" during the last two days (for comparison, there were eleven others closed for entirely other reasons during the same period).
That percentage does seem higher than maybe it should, so let's look at the questions more closely, one at a time, and decide whether that's valid for each:
Maths bug in collision response code. The asker provides a big block of code to sift through, including reading keyboard input and etc. On the plus side, the question provides a clear explanation of exactly what is going wrong.
I'm not a huge fan of big code-dumps, personally. But this one at least is reasonably well formatted and it's easy to skip past the obviously irrelevant bits. I guess my concern is that fundamentally, this is a "find the math bug in my code" question, rather than one which addresses a concept which can be explained or a technique which can be learned. See my comments re: discoverability of "solutions to a math bug" questions under the "arcball" question, below.
Has an accepted answer.
This is another "help me my code isn't working and I don't know why" post. Which probably isn't tremendously surprising as it involves quaternion math, which isn't a trivial thing for anybody.
The question is a huge code dump, and includes a download link to get even more code. If this had focused down on "what is the right way to do quaternion math", then I think it could have been a really good and useful question (if there isn't already such a question), but as is, I have trouble mustering enthusiasm to go delving through somebody else's math code just to go searching for an arithmetic bug.
I'm not certain how finding this guy's maths bug and indexing the particular located bug in Google helps the Internet as a whole; the next guy will presumably have a different maths bug, and we'll just wind up answering this same question again and again, just with different swapped variables or subtracts that should have been adds in each new asker's version of the code. So I'm inclined to agree with the "too localized" evaluation. Again, my view is that this isn't somebody asking for something to be explained, or to be taught; he's just asking us to find his maths error for him.
Maybe that view is a personal failing of mine.
I have trouble parsing the question, here. It appears to be another plea for someone to debug their code for them. Whatever the question actually is, even if someone was motivated to go delving through source code to find the bug, the question is fundamentally unanswerable, because not enough code is provided. My instinct is that this question, if pursued, would have turned into a "teach me how to program" question.
But I wouldn't have voted "too localized" on this question. I'd have voted "not a real question", myself.
This one doesn't look terrible to me. At least the problem is well-defined, and the source code has been trimmed down to a somewhat reasonable amount for people to peruse. Fundamentally, it's still a "I haven't checked for existing learning resources" sort of question, but I think this one could have been salvaged.
The real question from my point of view is whether someone else would ever have found it -- the asker's misunderstanding was regarding the precise usage of glBindBuffer(), which is fair enough. But none of that shows up in the question text -- there's no way for another user to find it. Once the real solution was found, I wouldn't have been against someone editing the question to really tighten it up to fit the actual problem, to aid in discoverability.
I suspect that this one was closed due to its very low prose:code ratio, which is usually a strong indicator of questions which are only of use to one person at one moment in time.
Has an accepted answer.
I'm having trouble understanding this question. I'd have asked the poster to rephrase in an edit, and voted "Not a real question" in the meantime.
This one doesn't describe the problem and doesn't explain what needs to be fixed. The question appears to be "Is it necessary to have a diagonal collision for the wall?", which there's no way for us to know without a lot more context (what the game is, how big the wall is, where it is, how thick it is, how far game objects move in a frame, etc). I think I'd have gone with "Not a real question" on this one as well; I don't think the question is defined enough to be "too localized". (Although I can see an argument that if nobody else can understand the question, its answer is probably not going to be of much use to them, either.)
Just from my examination of the last 48 hours, I guess I'm not really seeing a problem. Granted that there are some questions which I wouldn't myself have classed as "Too Localized", but many of them I would have voted to close for other reasons.
I was a little surprised that there were only these six "too localized" questions in the last two days -- I've had the same sense as the OP that there have been lots and lots of extremely localized questions being closed on the site, but that doesn't seem to be borne out by this data. Having only six closed in two days surprises me a lot. And further investigation suggests that this rate doesn't seem to be unusual -- there were only six closed for that reason in the previous two days, as well. And even fewer in the two days before that.
It's also interesting that both of the questions which I considered to be potentially answerable did get answers, and the original poster of each accepted those answers, even though the questions were later closed as "too localized". So the whole "too localized" thing doesn't seem to actually be stopping people from providing personal assistance when the questions asked are truly answerable. (as per my admittedly subjective opinion re: answerability).
Maybe a wider analysis of more data would show a different trend, but just going on the basis of the data from the last 48 hours, I'm not seeing anything particularly alarming in regards to questions closed as "too localized".