TL;DR: Sui's Mainnet Tunnel campaign was verified by CertiK, which found 15,241,348,297 committed tunnel transactions across 7,680 tunnels, with a peak of 40,682,328 TPS and a sustained 40,339,143.65 TPS. The key issue is not raw speed alone, but whether off-chain activity can remain auditable while entry and settlement stay anchored on chain.
At a glance
What this is: CertiK independently verified Sui's Tunnel performance campaign, confirming very high off-chain throughput with on-chain opening and settlement anchored to the mainnet.
Why it matters: Identity and access teams should read this as a governance pattern for high-volume, stateful systems: control must shift from every action to tunnel admission, authorization, and final settlement.
Context
Sui Tunnels are a state-channel style design that moves repeated interactions off-chain while keeping opening, custody, and final settlement anchored on chain. The governance question is whether the system can preserve verifiability and authorization when most activity is no longer individually written to the base layer.
This matters wherever large volumes of short-lived, stateful interactions need to be settled without turning the underlying chain into the bottleneck. The article is primarily about performance verification, but it also speaks to trust boundaries, admission control, and the evidence needed to audit off-chain execution.
For identity and access practitioners, the relevant lesson is that high-throughput architectures concentrate security at lifecycle events rather than every transaction. That is a familiar pattern in non-human identity governance, where the control plane matters more than the volume of routine actions.
Key questions
Q: How should teams govern off-chain execution when settlement stays on chain?
A: Teams should treat off-chain execution as a bounded lifecycle with explicit admission, signed state transitions, and verifiable closure. The control objective is not to inspect every action at runtime, but to make sure the system can prove who entered, what rules applied, and what final state was settled. That shifts governance to identity, authorization, and auditability at the boundaries.
Q: Why do lifecycle controls matter more than per-transaction controls in tunnel systems?
A: Because once repeated actions are moved off-chain, the base layer no longer sees each interaction. Security then depends on whether opening, custody, and settlement are tightly controlled and independently verifiable. If those lifecycle events are weak, the system can process large volumes quickly while still leaving gaps in accountability, reconciliation, or dispute resolution.
Q: What are the signs that an off-chain throughput benchmark is being overstated?
A: Watch for unclear transaction definitions, per-channel rates that are simply added together, or time windows that include partial seconds and edge effects. A reliable benchmark states exactly what counts as a transaction, which intervals are included, and how parallel workloads are reconciled. If those rules are vague, the number may be impressive but not decision-grade.
A: They should compare lifecycle assurance, not just throughput. A fully on-chain model keeps every state change visible, while a state-channel model concentrates trust at entry, custody, and settlement. The right choice depends on whether the workload can tolerate off-chain execution while still preserving replayability, audit evidence, and final settlement integrity.
Technical breakdown
How tunnel admission and settlement anchor off-chain activity
A tunnel is a bounded execution environment in which participants exchange signed transition receipts under agreed protocol rules. The chain records the opening state, the participants, their authorization keys, custody terms, and the final settlement, while the intermediate actions remain off-chain. That design reduces on-chain load, but it only works if the transition receipts can be replayed against the protocol rules and reconciled to the final on-chain record. In other words, the chain becomes the source of truth for lifecycle boundaries, not every state change.
Practical implication: Treat admission, authorization, and settlement as the primary control points because they define the trust boundary for the whole tunnel.
Why measured throughput depends on methodology, not just raw counts
Throughput claims can be misleading when parallel channels are counted incorrectly. Here, the measurement used qualifying tunnel transactions inside a defined play window, counted only complete one-second UTC intervals for peak and sustained figures, and avoided adding per-tunnel rates in a way that would inflate totals. That matters because off-chain systems can look faster than they are if the numerator and denominator are not aligned. A valid benchmark needs reproducible accounting, not just a large aggregate number.
Practical implication: Validate benchmark methodology before using throughput claims in architecture or procurement decisions.
Why shared lifecycle transactions become the bottleneck
The article shows that 15.2 billion off-chain transactions were anchored through only 1,920 on-chain lifecycle transactions. That means the system's practical limit is not just how quickly participants can act inside the tunnel, but how efficiently the base layer can open, fund, and settle those tunnels. For applications with many participants and short-lived sessions, the control challenge shifts to lifecycle orchestration, not individual transaction processing. This is the same pattern seen in other high-volume identity and workload systems where batch lifecycle events dominate risk.
Practical implication: Design capacity and resilience around opening and settlement rates, not only around in-session transaction volume.
NHI Mgmt Group analysis
High-throughput state channels are a governance problem before they are a performance problem: once activity moves off-chain, the control surface collapses toward admission, custody, and final settlement. The security question becomes whether those lifecycle events are tightly bound to authorization evidence and replayable records. That makes tunnel governance closer to non-human identity lifecycle control than to ordinary transaction processing.
Verification discipline matters more when a benchmark is spectacular: a result in the tens of millions of TPS is only useful if the arithmetic, timestamps, and reconciliation logic are independently checkable. CertiK's role here illustrates the importance of separating performance claims from assurance claims. Practitioners should treat measurement method as part of the security posture, not as a reporting footnote.
Bounded execution is a useful pattern for agentic and interactive systems: the article's tunnel model resembles how many autonomous or semi-autonomous workflows will likely be governed, with a controlled entry point, signed transitions, and final settlement. The named concept here is lifecycle-anchored execution: most actions occur outside the base ledger, but the system remains accountable because entry and exit are fixed and auditable. That is the right mental model for future non-human identity and agent governance.
Anchoring, not raw throughput, is what makes scale credible: the campaign's value lies in showing that large volumes of off-chain activity can be reconciled to a relatively small number of on-chain lifecycle events. This does not remove governance pressure, it relocates it. The more the workload scales, the more important it becomes to prove that the admission, custody, and settlement records are complete and consistent.
State-channel systems will increasingly converge with identity controls: the tunnel participant set, authorization keys, and custody terms are identity artefacts, even when the article is framed as performance engineering. That means identity teams should read these architectures as lifecycle systems with cryptographic enforcement, not as pure throughput infrastructure. The practitioner conclusion is to govern the control plane as carefully as the transaction plane.
What this signals
Lifecycle-anchored execution: the useful security boundary in high-throughput off-chain systems is not each individual action, but the governed entry, exit, and reconciliation points. That is the same architectural lesson identity teams see in non-human identity lifecycles, where control at issuance and offboarding matters more than reviewing every routine call.
As more systems push repeated interactions off the base layer, practitioners should expect assurance to shift toward proof of admission, settlement integrity, and replayable evidence. The practical test is whether the programme can still explain who was allowed in, what rules applied, and what was finally committed.
For practitioners
- Map admission controls to lifecycle boundaries Define who or what may enter a tunnel, what authorization evidence is required, and which records must exist before settlement is accepted.
- Separate benchmark arithmetic from assurance claims Review how qualifying transactions, time windows, and concurrency are counted so that throughput figures cannot be inflated by parallel-channel math.
- Treat tunnel keys as governed identity artefacts Inventory the authorization keys, custody terms, and protocol rules attached to each tunnel and manage them like short-lived machine credentials.
- Design for opening and settlement bottlenecks Size capacity around the rate at which tunnels can be opened, funded, and closed, because those lifecycle events become the limiting control points.
Key takeaways
- The article shows that very high off-chain throughput is compatible with anchored settlement, but only if lifecycle controls remain tight.
- The assurance value of the benchmark comes from reconciliation, defined measurement windows, and independently checkable evidence.
- Practitioners should focus on admission, custody, and settlement governance when workloads move away from per-transaction on-chain processing.
Standards & Framework Alignment
This section maps relevant standards and security frameworks to the operational risks and controls described in this guidance.
MITRE ATT&CK addresses the attack and risk surface, while NIST CSF 2.0, NIST SP 800-53 Rev 5 and CSA Cloud Controls Matrix set the governance and control requirements practitioners need to meet.
| Framework | Control / Reference | Relevance |
|---|---|---|
| NIST CSF 2.0 | PR.AA-05 — Access Permissions, Entitlements and Authorizations | Tunnel admission depends on controlled authorization keys and settled entitlements. |
| Recommendation — Apply PR.AA-05 to govern who or what may enter a tunnel and what evidence authorizes settlement. | ||
| NIST SP 800-53 Rev 5 | IA-5 — Authenticator Management | The article centers on authorization keys and lifecycle control for repeated signed actions. |
| Recommendation — Use IA-5 to manage tunnel authenticators, rotation, and revocation across lifecycle events. | ||
| CSA Cloud Controls Matrix | IAM — Identity and Access Management | Tunnel participants, keys, and custody terms are identity controls in a cloud-style execution model. |
| Recommendation — Map tunnel participant and key governance to IAM controls for admission, custody, and closure. | ||
| MITRE ATT&CK | TA0006 — Credential Access | The tunnel model depends on signed authorization evidence that must not be abused or replayed. |
| Recommendation — Hunt for credential and authorization abuse around tunnel admission and settlement records. | ||
Key terms
- State Channel: A state channel is an off-chain execution pattern where participants exchange signed state transitions and later settle the final result on chain. It reduces base-layer load while preserving a verifiable boundary for admission, custody, and closure.
- Tunnel Settlement: Tunnel settlement is the on-chain act that records the final agreed state after off-chain activity has completed. It is the governance anchor that lets participants prove what happened without writing every intermediate action to the ledger.
- Transition Receipt: A transition receipt is a signed record that represents one valid step inside an off-chain execution flow. In this model, the receipt is the evidence used to replay, reconcile, and verify the final state against the on-chain record.
- Lifecycle Anchoring: Lifecycle anchoring is the practice of binding admission, custody, and closure to an authoritative ledger while allowing intermediate work to happen elsewhere. It shifts assurance from every event to the control points that define who was allowed in and what was finally committed.
Deepen your knowledge
NHI Foundation Level course, the industry's only accredited NHI security programme, covers NHI governance, machine identity security, and identity lifecycle control. It is designed for practitioners who need to govern short-lived, high-volume access patterns across modern security programmes.
Published by the NHIMG editorial team on October 16, 2026.
NHI Mgmt Group, the independent authority on Non-Human Identity, IAM, and Agentic AI security. nhimg.org