Avalanche: who orders transactions?

Control profile

Transaction ordering

599 Primary Network validators, top validator 1.72% of stake; no single operator

Escape hatch

n/a — L1

Upgrade authority

No admin key over user funds. Upgrades run through the open ACP process, but in practice Ava Labs writes, releases, and schedules every fork.

Liveness record

Two full outages, both single-client bugs fixed by an emergency Ava Labs release: 2023-03-23 (AvalancheGo v1.9.12) and 2024-02-23 (~5h finality stall). No rollback, no fund loss, and no halt found in the ~2.5 years since, the same recency as Solana's record, without Solana's second client on the horizon. Validator count has roughly halved since 2022 (~1,200 to 599), coinciding with Etna/ACP-77 letting L1 operators stop validating the Primary Network. Nakamoto coefficient 23, measured on-chain at this review.

Ketju control assessment

UPGRADED from under-review, and the question that blocked it dissolves on inspection: the C-Chain, where every venue we track lives, sits on the Primary Network and inherits nothing from subnets. A subnet failure is an economic event (smaller security budget for that L1), not a consensus event for the C-Chain. What remains is a genuinely decentralised ordering layer, 599 validators, Nakamoto coefficient 23, no admin key over funds, no freeze mechanism demonstrated or designed, carried by a concentrated supply chain: AvalancheGo is the only production client, both historical outages were single-client bugs waiting on one vendor's patch, Messari reports over 33% of stake hosted on AWS, and Ava Labs runs the default RPC, indexer, and wallet pipeline. A compelled Ava Labs could not take funds, but it could ship a censoring client or filter the default RPC, and the network's history says recovery from any halt runs through one company. Approved with limits reflects that split exactly, a split of the same shape we priced on Solana: sovereign-grade fund control on one axis, a concentrated operating layer on the other. This unblocks benqi-staked-avax and the other Avalanche-only venues to be judged on their own merits.

Observable review triggers

• Any full-network halt requiring a coordinated restart
• Nakamoto coefficient falling below 15
• Primary Network validator count falling below 450
• A single hosting provider exceeding 40% of stake
• The Avalanche Bridge wardens (6-of-8 as last published, 2022; Ava Labs-selected) moving collateral outside the published process

These triggers make the assessment monitorable. They are not predictions; each identifies a fact that would change or reopen the current control assessment.

Advisor implementation questions

1. Does the client’s thesis require censorship resistance, or primarily low-cost settlement?
2. Can the client exit without cooperation from the normal transaction-ordering party?
3. Who can upgrade bridges or contracts, and what delay applies?
4. Does the asset introduce an issuer weaker than the chain grade?
5. What evidence will show a halt, censorship event, or governance change?