Robbo,
My understanding is that they do it in the following manner:
Firstly, the unstable moment created by the CG being out of line with the pivot point is balanced by hub moments. This is inherently dependant on the pilot/RC Operator having the reactions to correct the divergence before the body rotates to such a degree that the maximum moment achievable by the hub is exceeded and the aircraft crashes.
I agree with your analogy with the up-turned pendulum, this configuration is statically unstable - but that's not to say it couldn't be designed to work if the need arose. Blades don't need to be rigid and theoretically any hub will work though the more control power the better!
CRAN