If you know a better way to deal with malicious nodes, it would be useful to point out a specific method that you think works in this setting. We can then discuss it.
To be concrete, I briefly looked at the paper “Resilient Network Coding in the Presence of Byzantine Adversaries”. The paper shows among other results that if an adversary can inject z packages per time and the network capacity is C, then the code achieves a rate of C - 2z. This means it is assumed that a receiver obtains more honest packages than malicious ones. This does not seem applicable in the blockchain setting, where honest nodes try to minimize their network traffic (instead of exhausting the network capacity) and adversaries can spam the network. Note that the Pedersen commitment approach does not suffer from this (but instead assumes a computationally bounded adversary).