It's likely because of timeouts configured in your server for document reloads which, by default, it's set to 21600 seconds (6 hours). You can check that in your Management Console "User Documents" -> "Reload". Try with some higher values and see if that helps.
That's just a warning, meaning the file has not been found in the path you provided. But the file could be created again. Actually, the system itself "skips" that. That's happening to me (sometimes) but the reload process finishes correctly.
Anyway, I'd try debugging manually instead of reloading the whole document, limiting load to a few hundreds of records, just to see step by step what problem is knocking off your system. Take into account that if extraction process takes up to 7 hours, amount of record must be huge, so memory would be affected by any other processes running in that computer... May it be something related oto your hardware requirements for a so big amount of data?