That 1cm accuracy and degradation with distance refers to a local coordinate frame, Earth-centered now but probably will be sun-centered. But this coordinate frame can be nested within another, global one, with say 2 billion resolution. That way you can cover 2 billion * 2 billion km with 1cm resolution ...
You could take sun as an absolutely fixed object, divide space around solar system in 100 000 000 km sized cubes (one sides length), and then compute coordinates relative to the current cube you're located in. Maybe cube size needs to be a bit smaller... but you'll never run out of precision in the interplanetary flight.
When approaching some body switching to body-relative coordinate system is totally justified, and even though it's non-inertial, the perturbations caused by the body you approach will greatly exceed the error in computations.
When flying through interplanetary space, integrating within a single "cube" is possible without losing precision at any point. Switching between these coordinate system is a matter of trivial add/subtract. It doesn't even have to be known to the graphics engine - you will never notice the error in the interplanetary space (no reference bodies other than yourself), and your coordinates will be relative to nearest body once you're there (so back to high precision again).
This might seem a bit pointless (why not make just one double-precision space?), but implementing this is trivial, and it could be used to build interplanetary flight simulation with great great precision.
I'm not really planning to making interplanetary flight sim just yet, but if the outerra engine is nice and smooth enough (for my aerospace simulator), I would start work on that (the physics side of things).
Really looking forward to the wiki, would be interesting to see and work on the API for addons.