The thing with math (and abstract CS) is that it's generally purely logic. In principal you should be able to write (at least verify) an entire article with no outside references as it is all logical deductions, it's armchair science in that sense (no negative connotations meant).
But if you're going to write an article on say quantum computing or confocal microscopy, to name random things, you are going to need outside references to experimental work, history, best practices, open debates, etc. And that is usually where the errors slip in, how many people actually check the sources properly?
They're also usually pretty terrible when it comes to modern science things, but I suppose that can be attributed to a small group of people actually knowing about those things in the first place.
This is for a very narrow definition of maths. Also it's rare that a paper is completely comprehensive. Some proofs are usually defered to other papers.
But if you're going to write an article on say quantum computing or confocal microscopy, to name random things, you are going to need outside references to experimental work, history, best practices, open debates, etc. And that is usually where the errors slip in, how many people actually check the sources properly?
They're also usually pretty terrible when it comes to modern science things, but I suppose that can be attributed to a small group of people actually knowing about those things in the first place.