Obviousness requires a reputable, dated source stating something like "I have combined these two elements" or "Someone should combine these two elements". While magazines or trade literature are acceptable sources, the prior art search is typically conducted primarily through the patent archive. It's a slam dunk if you have two patents with claims that cover the same invention and some way in the text to tie them together. You can't generally rely on your 'expert opinion' that it's obvious because you say so.
Patent examiners are also promoted and retained primarily based on the volume of patents they process. IIRC the bar is something like one or two patents processed a day. That would mean 6-8 hours of reading the patent, researching prior art, writing the response, reading the attorney's response and writing your final disposition, as well as any appeals. Regardless of size of patent.
All this for mid 5 figures to low six figures in NOVA
My guess is that it's not necessarily that people at the patent office have a hard time telling if something is obvious, so much as that the incentive structures at play are all wrong.
I imagine that, if a patent reviewer denies a patent application, they potentially have a fight on their hands when their decision gets challenged. Whereas, if they approve it, any argument ensuing from that decision will be the court's problem. It's a classic "keep your head down and don't cause any trouble for yourself" situation.
Assuming I'm right on that guess, I'd further guess that it's also a classic "I'm not paid enough for this shit" situation. I don't have any friends at USPTO, but I have friends who work in other government offices, and, from what they've said, I would assume that that is a deciding factor in a large percentage of decisions that are being made by rank-and-file government employees across all branches. And I doubt patent clerks are any more likely than any other kind of clerk to be getting paid "dealing with this shit" money.