But how big a weakness is it, assuming there's one at all? True there aren't any diodes on the inputs, but there are other protection measures in place that the AES Phantom Menace paper recommends.
First, the paper assumes 10 Ohm resistors on the input of the instrument amp for the simulated circuit and then goes on to use 5 Ohms on their actual test circuit. The DMP3 has 47 Ohm resisters (or just shy of half of what they state would be the safe minimum for their simulated circuit in part 1.2). So potentially just a little more than 1/10th the maximum initial current hitting the inputs compared to the test circuit. The fact that the total charge is distributed between two capacitors in series may reduce the effective current further still. I'll leave it to someone with more hands-on as opposed to theoretical electronics knowledge than me to determine if that "may" is a "does" or "does not".
Second, the paper assumes 47uF capacitors on the input; the DMP3 has two 10uF in series for a total of 5uF. So the potential stored charge on the phantom blocking caps is 1/10th as much as the value calculated in the paper and a lot less stored charge to cause damage to the chip.
It's hard to say whether these mitigations alone are enough to adequately protect the INA163, but on the other hand a search--albeit an admittedly quick and not thorough one--didn't reveal any widespread failures among users. The Rane on the other hand has the same values as specified in the paper (10 Ohm resisters, 47uF caps) and probably does need the protection diodes. All that being said, if you wanted to add some diodes to the front end, go for it; I doubt it would cause any issues and could very well save your bacon if you accidentally hot-plug something.