Re-reading 310.4, I believe that you are correct, but believe that violating 310.4 in this case would result in a safe installation. It seems to be 'common sense' (which is often wrong!!) that it should be okay to have parallel conductors where each _individual_ conductor could safely carry the entire load. In fact Exception 2 of 310.4 points to this very condition, though it is not applicable because it is for control instruments.
Additionally, one could argue that while these conductors are _electrically_ in parallel, meaning that some current will flow in each path, the fact that these conductors follow different paths through different components (different switches, etc) that these are in fact _not_ paralleled conductors. I agree that this is a _very_ weak argument
IMHO there should be an exception for 'switched branch circuit conductors in parallel where the _individual_ conductors are capable of supplying the load, and properly protected by the branch circuit OCPD.'
The normal rules about paralleling large conductors are there to make certain that the load is properly balanced between them, and that any variance in the balance will not damage one of the conductors.
If the conductors were arranged such that current in each of the parallel paths was restricted to the individual cable, to get rid of the current loops, then I believe that the installation would be safe, but as Bob points out the parallel conductor rule makes it a code violation.
An alternative approach would be to wire a single pole switch loop to short the two traveller terminals of one of the three way switches together. This would also result in a 'parallel path' situation, but it is now limited to the current flow path _inside the switch_ being in parallel with the entire switch loop.