Ossification Kills Decentralization - New Paper Redefines Trust and Incentives

Measurement studies keep showing blockchains are less decentralized than advertised, but the usual suspects – skewed token distribution, mining pools – are symptoms, not the root cause. A new arXiv preprint (2606.29826) from the Ethereum research community pins the blame on a far more insidious force: ossification.

Ossification is what happens when a coalition of nodes over time pools resources, collaborates actively, improves efficiency, and slowly dominates the marketplace until smaller players can’t compete. Mining pools are the textbook example. The authors define a decentralized system as one where there is no incentive for any subset of nodes to ossify. By that standard, every major blockchain today fails.

Shapley Values Over Collaboration DAGs

The paper’s core mechanism replaces competitive block rewards with collaborative rewards computed from a rooted directed acyclic graph (DAG) of collaborations. Each node earns an “importance score” based on how many diverse collaborators it has worked with over time. Rewards are allocated using an asymmetric Shapley value with DAG-dependent weights that provide Sybil resistance – you can’t pretend to collaborate with others when it’s really just yourself.

Crucially, collaboration choices are entirely subjective. Nodes pick who they work with, and the system penalizes static collusion sets. A group that only collaborates among themselves sees diminishing rewards. This forces even large players to seek out smaller, diverse partners to keep their importance scores high.

Trustless No More – Trust Is the Point

This is a marked departure from the trustless ideal. Instead of algorithmic committee assignments (like sharding), the protocol explicitly slashes the collaboration initiator if any partner misbehaves. Known miscreants will be voluntarily excluded because honest nodes won’t pick them. Changing identities doesn’t help because importance is earned over time and cannot be transferred.

The authors admit that efficiency-heavy services – cloud, high-throughput payments – are probably unsuitable for an unossified blockchain. But services that benefit from human trust, creativity, and diverse expert collaboration (think a decentralized medical board) thrive under these incentives. The goal isn’t to beat AWS on latency; it’s to offer transparency and radical inclusivity that centralized platforms can’t replicate.

Scalability falls out as a side effect – collaborative block production naturally increases throughput without forced shard assignments. The open question is whether human nature will re-introduce leadership hierarchies, or whether the mechanism is strong enough to keep the system genuinely flat.

Source: Rethinking Collaborative Trust for Verifiably Decentralized Blockchain Systems
Domain: ethresear.ch