I think the two points that would be put forward in favor of this approach are:
1. Reduced resistance due to the parallel path.
2. One bad connection will still leave a good ground via the other side of the loop.
I know the NEC has limitations on paralleled conductors, but I'm guessing that this would not apply to grounding electrode conductors as they are not designed to carry current under normal operation. (And there are already plenty of parallel paths due to all the bonding anyway.)
Maybe our code experts can confirm?
