I want to know the technical details of what Cummins allegedly did "enough that they're not arguing a massive fine, while claiming they didn't do it on purpose."
Claiming they "installed defeat devices" isn't nearly enough detail. How did it alter either the engine combustion cycle or the emissions control system behavior?
Who knows? I believe VW chief engineers got in jail not knowing what their subordinates did. I really tried to understand the VW scandal but I could not.
When operating an engine (gasoline or diesel), you can make a range of decisions about how you run it. Fuel/air ratio, ignition timing, exhaust gas recirculation amounts, boost pressure, etc. There are a lot of ways to accomplish a given power output, and generally some will be more or less efficient in terms of fuel quantity used, and some will be better or worse for emissions. You'll usually find that "optimum for one" is not "optimum for all," and so the ECU (engine computer) has to decide where to run the engine for any given situation.
The (increasingly invalid) assumption made by the EPA test cycles for emissions is that "What it does on the test treadmill will generally reflect what it does on the road." So they've got some (well known) set of drive cycles that reflect a range of driving conditions, they put a car through its paces on the treadmill, and this (a) validates that the car meets the relevant emissions standards, and (b) gets you the EPA-rated miles per gallon. Probably other stuff too, but these are the important ones.
VW violated, quite intentionally, the "behaves the same in the test chamber as on the road" assumption. Every time the car was started, it would assume it was being tested for emissions, until something "impossible in a test chamber" happened (like the steering wheel was turned - you don't do that on rolling road treadmills). At that point, once it was confident it wasn't being tested, it would alter the engine maps to be rather a good bit more fuel efficient - but at the cost of NOx (oxides of nitrogen) emissions being also far higher. So, on the road, the result was that the cars beat their EPA rated fuel economy (which is odd - nobody ever does that, as the real world has things like wind), but also violated the emissions regulations they were certified under (while also putting out less CO2 - because it used less fuel than rated for those conditions).
Most likely they've been caught with memos saying to install a device that would trick past the test. It doesn't really matter what the devices did on an engineering level or even whether they worked at all, and the regulator may not even know or care; the hard part in a case like this tends to be showing criminal intent.
Similarly you could lie on the OC2 response.