I know what your trying to say but this isn't the same kind of layering you are talking about. Look at it yourself.
I see what you are saying when it comes to weird layers. All I can say is that his birth certificate looks a lot like the copies of old engineering drawings we have on file. Back before computers, all engineering drawings were done on big sheets of paper. When computers became practical, they started to scan these drawings as high resolution raster images. They then make a new drafting file with the raster as a background. As we changed or added things to the plan, we would clip out the raster portion and draw in vector graphics. The result is a weird combination of both vector and raster. The raster part looks just like a scanned document; it's a bit fuzzy. The vector stuff looks perfect.
Obama's birth certificate has obviously been passed through an OCR because you can see portions of a word will be recognized as text and redrawn and other portions in that same word are part of the background which is not redrawn. Go to the part where it says birth place of mother. The K in Kansas is original raster and it looks genuinely terrible. The "ansas" part was picked up by the OCR and drawn to look terrible just so it matches the rest of the word. Adobe Acrobat does that if you set it to OCR then downsample. The other options available are to OCR but don't change anything (I use this on official scanned pdf documents) or OCR then redraw as proper text.
I think only one of the signatures looks strange. The local registrar is almost entirely raster as it should be, and attendant is raster as well. Signature of parent seems to be part of the OCR recognized text even though it doesn't look like the type of text that can be read with OCR. “Ann D” appears to be part of the background raster, but “unham Obama” is drawn in separately when zooming.
You asked why it would have so many layers. That's just part of how Acrobat optimizes things. If you scan something, it will try to pick apart objects and save them independently. When I use software to convert a scanned PDF document to a rough approximation for AutoCAD, the software will spit out multiple files. Each of those files it created for AutoCAD was an image layer in the scanned document. Sometimes it doesn't even make sense how it would split up the drawings. Something like a title block around a page will be split up into multiple files because it was recognized as multiple independent images.
Well why they wouldn't flatten it is beyond me in any case.
Those multiple image layer PDF files I mentioned are generated by the office Xerox copier. I don't make them on my computer. If the people scanning that page have the same copier I have at my work, then that's why it isn't flat. Something about having multiple images makes it easier to compress somehow, or the compression is more effective.