It's not just the odd flubbed line, it's loads of problems. I hear a lot of stories about it. Sometimes the narrator's pronunciation of a weird name can drift, very gradually, and the engineer then has to go a long way back to figure out where the drift started, and get them to restart from there. It has to sound natural and flowing, so no you can't just re-record individual lines and splice them in later, they would stand out and it would sound shit. That's why it can take up to a few hours to get one hour of audio down. (But not 5-12 hours, he says that's a huge exaggeration.)