In the 90's Computer Companies would test just over 80% of software code, as it was just too expensive to test higher than that. By using the computer for all sorts of applications, the Customers would be testing the remaining 20% of code and discover any "bugs" that were still there. Once the bug was discovered by the Customers, a software update or patch would be sent out to all Customers. Thank goodness for all concerned that mainframe computers don't fly very high.
From what Thor Nogson says, if the source code has 8 million lines of code and if the same rational is still used, then there could still be over 1 million lines of code that has been untested and have bugs in the code.