Control profile
- Transaction ordering
- Leader schedule across the validator set — no single operator
- Escape hatch
- n/a — L1, no exit dependency
- Upgrade authority
- No admin key over user funds. Client software upgrades are social.
- Liveness record
- Last confirmed major halt: 2024-02-06, a ~5 hour outage caused by a program-execution bug. Since then: 22+ months of uptime and 100% year-to-date as of March 2026. Frankendancer (Firedancer networking + Agave execution) has run on mainnet since 2025; once Firedancer passes ~25-30% of stake, a single client bug can no longer halt the network. Concentration is moving the other way: validators fell ~2,560 (Mar 2023) to ~795 (Jan 2026), a 68% decline, and the Nakamoto coefficient fell 31 to 20. Validators remain spread across 45+ countries and multiple hosting providers.
Ketju assessment
UPGRADED from under-review. The earlier memo leaned on "has halted and required coordinated restart" without pinning the date — and the date turns out to matter a great deal. The last major outage was February 2024, nearly two and a half years ago, and the structural fix is not a promise but shipped: multi-client execution is live on mainnet. Twenty-two-plus months of uptime is a real record, not a lucky streak, and holding the old verdict would have been stale caution rather than judgement. Liveness and decentralisation are moving in OPPOSITE directions here, and they are different axes. Liveness has improved structurally; validator concentration has worsened materially. Approved-with-limits reflects exactly that split — the chain is usable, the trend in operator count is the thing to watch, and no admin key can seize client funds either way. This unblocks Kamino, Jito and the Jupiter venues to be judged on their OWN merits rather than deferred behind the chain.
Observable review triggers
- Any full-network halt requiring coordinated restart
- Nakamoto coefficient falling below 15
- Validator count falling below 600
- Firedancer adoption stalling below 25% of stake
These triggers make the judgment monitorable. They are not predictions; each identifies a fact that would revoke or force review of the current assessment.
Advisor implementation questions
- Does the client’s thesis require censorship resistance, or primarily low-cost settlement?
- Can the client exit without cooperation from the normal transaction-ordering party?
- Who can upgrade bridges or contracts, and what delay applies?
- Does the asset introduce an issuer weaker than the chain grade?
- What evidence will show a halt, censorship event, or governance change?