Money requires consensus. I think it's the FLP Impossibility result that ultimately prevents CRDT-type (eventually consistent) schemes from implementing consensus.

Researchers in eventual consistency pretty much just assume this.

e.g. in the original CRDT papers -- bottom of page 10 in : https://arxiv.org/pdf/1805.06358.pdf; top of page 11 in : https://hal.inria.fr/inria-00609399v1/document