Can you take one more evolution question? ;-]
How do you get a gradual mutation to do this?
Disclaimer: I haven't read the papers describing the technical details of this specific process, so I'm making my best educated guess based on what I already know about bacterial molecular biology. I'm not going to go to the beginning and describe a scheme by which all of the separate genes evolved; I'll just start at the part where these two pathways became linked.
What it looks like to me is that the bacteria had two "expression cassettes" located close to each other on the chromosome.
An expression cassette is a set of genes and associated regulatory elements that are all involved in a common pathway, and are placed next to each other on the chromosome. A bacterium only has one chromosome. Expression cassettes are very common in bacteria (and, to my knowledge, are not used by eukaryotes [organisms with nucleated cells, which includes all multicellular organisms]).
All organisms have a method of moving pieces of DNA from one part of a chromosome to another. These pieces of DNA are called "transposons." Some transposons require an enzyme in order to move. Transposons typically have the same, or complementary, DNA sequence at each end, which is necessary for them to be able to move. I'll call these identical or complementary DNA sequences "sticky ends."
What I surmise happened in these bacteria is that the control elements--promoters--for each of the two expression cassettes became part of the same transposon. This could be from random mutations, or it could be because a transposon inserted itself there and left behind copies of its sticky ends when it left. In any case, once the promoters become a transposon by being between sticky ends, all they have to do is excise themselves, turn around, and reinsert themselves. It's kind of like splicing a film, if you want a visual with that--you can cut out a segment of film, turn the cut piece over, and reinsert it backwards, and it will still run through the projector because the little holes on the edges still line up. Once that promoter area becomes a transposon, the bacteria no longer needs two promoters, so if one is deleted (through a "glitch" in transposon activity), the cassettes will still work... just not both at the same time, because there is only one promoter now. At this point, bacteria that can control the switch--probably through a transposase (the enzyme that catalyzes some transposon movements) mutation--have an advantage over bacteria that can't control the switch.
Switching mechanisms such as that described here are not unknown. Pathogenic bacteria use similar switches to evade the host immune system.