My employer's legacy system started off (in 1997) as very structured procedural code (with a smidgeon of OO), and it was luckily one of the 10% of software projects that was successful. It may look a little antequated today, but at least it is structured and provided a framework to hang extensions upon.
Fast forward to the late 2000's and developers not connected with the original architects have bit by bit extended the system in an ad hoc way with no architecture, just making decisions as they went along, going with what seemed to make sense to them at the time. They also abandoned any naming or coding conventions. To over-simplify, they never make a class, never make anything for re-use, and the only consideration is: "will it break what is already there?" which puts a damper on more innovative approaches. There is still lots of VB6 code.
These developers are still there, and they are entrenched, and can patch the system, or do things allow by the original architects, very fast.
However, since there is zero code reuse, they have no building blocks with which to build an entirely new system, so they can't do anything really special or novel for a prospective client, even a giant one like Sams Club.
I recall there were some really good arguments for improving code quality, for example, by using OO methodology, in money terms that would get an executive's attention, but I can't recall where I saw it. It has been a while. I intuitively sense that gathering all functionality into a class provides the most flexibility and maintainability but I don't have any ammo.
Some of the thoughts I'm up against include: The system seems to be working "just fine". So why should we rock the boat with something like a "quality program"? The current developers love it the way it is, so they don't want to do anything to improve their code or the quality of the architecture. They know every one of the 73 places a certain quantity is referenced. Who needs it? T-Sql doesn't need to be compiled! Copy and pasting the same code localizes bugs to that particular copy.
It was unclear what I was asking, huh? Well, the answer accepted contains a comment that has a link (http://blog.codinghorror.com/the-big-ball-of-mud-and-other-architectural-disasters/) that was exactly what I was looking for, so I marked as answer. So, someone, perhaps several people, inherently knew what I was asking anyway. If you read the article sited, you'll see that actually what I'm going thru here is common. I am very much not alone in this, which makes me feel better. :-)