An article I read about 8 years ago. I didn't save it. But I can explain how JCL works:
It is designed to run "batch" programs. Typically, programs designed to process records in large files. A good example would be a program at a bank that runs at night against all savings accounts, adding interest by computing the amount of money in the account against the interest rate. This is a simplification, but you get the idea.
The JCL is a very short program, as little as three or four lines, that tells the computer what input file to read, what output file to use, and what program to run. However it may include lines that contain a label and a value. For example, the value could be "A", or "B" or "C", and in the COBOL batch program, One of the first lines would be a simple piece of code asks the question:
If CODE = A, perform 'calculate-interest'
Else
If CODE = B, Perform 'subtract-interest'
Else
If CODE = C, Perform 'Add-1-million-dollars-to-my-account'
Else
Display "invalid code"
This is how a programmer could literally give himself a million dollars, because the paragraph would simply find a particular account number and then add one million dollars to it. And the program would end.
Which is why we always validate and peer review code.
BTW, I've not written or even read COBOL for ten years, so there may be some specifics missing there. But the point is that a program can have what looks like "junk code" but in reality it is only used when certain input values are entered. Like DNA.