GMP,
Mike's rule of thumb: 2-5% code change is manageable; 5-10% is a major change; more than 10% means it might be cheaper to rewrite the entire code from scratch than try and adapt that in place.
Are you the Mike? I believe there was an issue with the E3 software. Millions of lines of code and consisting of many modules. One team might latch on to a particular module for a particular function that had not originally been intended. The original purpose may have ceased and there would be a temptation to 'clean up' the software but this could not be done because of unknown interdependencies. Was that feasible?
If it was, ultimately you have a massive amount of unknown redundant software hogging storage and processor capacity but performing no useful function.
On the Mk 1 Nimrod it had only 8k of 16-bit storage. Even there, despite their best efforts as squeezing a quart out of a pint pot it was found that some essentially similar routines were duplicated - each weapon type had its own ballistic formula when all it needed was a set of discrete constance and common variables.