Actually there was a guy at one of the big think tanks (Bell Labs or something) that started reducing his arguments to really small steps, skipping nothing. He kind of set up subsets of the proof like subroutines, referring you to later pages to get the details, etc. He found that a lot of things he thought he had proven actually had holes in them. He went back and looked at his previous papers and found that some large fraction of them were wrong (I don't remember the number, something like 1/3). At first he thought that he had been fooling himself all along and that he was a terrible mathematician. Then he started applying the same analysis to other published work, and he found the same ratio.
It wasn't that the results were necessarily wrong. In other words, what they were claiming might still be true. It was just that they hadn't proven it. There were holes in the places they had glossed over with a "clearly" or "we know from X that Y".
Unfortunately I can't remember the name of the guy or find his paper now. (Look! I'm a crackpot!)
Anyone else remember reading that?