Ron:
That's not quite right. With two dice, there are 36 combinations you can roll. Of those combinations, 11 have at least one six in them. Not 12. You have 6 combinations that start with 6, and 6 combinations that end with six, but rolling 6 and 6 is common to both lists. If your engine's chances of failing are 1:1000, then if you assume that all engine failures are completely independent events that don't involve factors that affect the other engine as well, then your approximation is close to correct. But that's a questionable assumption. For instance, if the fuel systems are interconnected, you could just plain run out of gas, affecting both engines. If the systems are not interconnected, but you filled them both with the same bad gas, that will affect both systems too. If both have EFI or something and are connected to the airplane's electrical system, then they could both be brought down by the same electrical problem. It also seems like, if you forgot carburetor heat on one engine, you'd probably forget it on the other*. And so on. I'm not contesting that two engines makes the failure of at least one more likely, but it's not twice as likely.
I don't disagree that, for asymmetrical thrust and a pilot who's not sufficiently and recently trained for it, the risk is increased rather than decreased. This seems especially problematic for an ultralight. Especially if Vmc is above stall. OTOH, if the aircraft can climb reasonably well and it's EASY to handle with one engine out, maybe the risk is reduced. Might be a tall order unless you have a push me pull you setup, i.e. both engines on the centerline. I seem to recall that the on the Cri Cri, if you lose an engine, the flow of the other one is supposed to stay attached to the canopy and gets directed at the rudder a bit crossways. I don't know how well this works. I've seen pictures of airplanes where the engines are canted outwards. I don't know how well that works either, and I'm sure it doesn't help efficiency. My guess is that centerline thrust might be the easiest to implement, but there might be an efficiency hit and cooling for the rear engine might be an issue.
I'll admit that the second paragraph above may be off a bit, but it's based on a lot of stuff I've read. The first is too, but I've read about a fair number of accidents and their causes.