Why editing text in a PDF is limited
A PDF stores where each glyph sits on the page, not the sentence it belongs to. That one fact explains every restriction you have run into.
What is actually stored in a PDF?
Positions. Each piece of text is an instruction to draw certain glyphs from a certain font at certain coordinates, and a page is a list of those instructions. There is no paragraph, no line break and often no space character at all: the gap between two words can be nothing more than the next instruction starting further along.
This is why the format is dependable. A PDF looks the same everywhere precisely because it fixes everything in place, and everything that makes it awkward to edit is that same property seen from the other side.
Why does the text not reflow?
Because there is nothing to reflow. Replacing a short word with a long one cannot push the rest of the line along, since the following words carry their own coordinates and will stay exactly where they are, overlapping the new text if it runs into them.
Real reflow would mean rebuilding the paragraph structure the file never recorded, then laying it out again with the original’s fonts, spacing and hyphenation. Tools that claim to do it are inferring that structure, and they are convincing right up until the moment they are not.
So what can an editor do?
Add and cover. New text, notes, highlights, shapes and images can be placed anywhere on the page, and an existing passage can be covered and retyped. That handles the ordinary jobs: correcting a date, filling a blank, blacking something out, adding a note before passing a document on.
It stops short of typography. Retyped text will match the original only when the font is embedded and available, and matching a scanned document is not possible at all, because there is no font in it to match.
When should I edit something else instead?
Whenever the source document still exists. Editing the file the PDF was exported from and exporting again takes less time than any amount of covering and retyping, and the result has no seams in it.
When the source is gone and the changes are substantial, extract the text and rebuild the document. That is honest about what is happening, and it leaves you with a file that can be edited again next year.
Where to go from here
Edit PDF places notes, shapes and text boxes on the page, PDF to text is the way out when the changes are too large to patch, and flattening fixes everything you added into the page content once you are done.
What OCR can and cannot read covers the scanned case, and PDF forms explained is the right page when the thing you are trying to change is a field.