The paper discusses this. The issues here are all idiosyncratic to Threema's design. I don't know what "classical architectures" refers to here, but one of Paterson's research areas is formal modeling of cryptography protocols, and a recurring complaint in this paper is that Threema is at turns either so simplistic or so weird that it is hard to apply the formal models the field has already built to it.
What issues are fixable, which are more endemic to the classical architectures?