I'll be doing the usual blog/forum post when Revision 3 goes live (later tonight or tomorrow), but there's an important enough breaking change around Box2D that I wanted to give everyone a heads up. (I'll reiterate all this in my release post.)
The original Box2D wrapper (it turns out) inconsistently applied Moai's unit scalar. I had hamfisted the unitsToMeters scalar in to make things easier for our designers, but I missed a bunch of places where it was needed. A lot of the force and torque setters, inertia and momentum for example. When we were working on Wolf Toss, we didn't think much about the physical accuracy of the simulation. We just treated the forces, etc. as magic numbers and tweaked them until the game felt right, so nobody really noticed that the numbers didn't make sense.
Enter 1.0 revision 3. Thomas Harning went through our binding with a fine-tooth comb and cleaned up a whole slew of my mistakes. Good news: the unit scalar is now much more robustly supported. Bad news: if you've been relying on the original implementation to make your physics work, your game is going to break.
There are two ways to work through this. Ideally, go through and modify your physics constants. Failing that, make a back-compat wrapper in Lua that applies the inverse of unitsToMeters on the parameters of all the methods that were fixed. We're taking the former approach internally for Wolf Toss. If someone wants to start in the latter it could be a community effort - probably no one out there is using *every* method in the old API.
Sorry for the inconvenience. This fix is very much for the best, even if it breaks us in the short term. Just wanted everyone to know about it ASAP.