Optimism Proposes Super Root Shift for OP Stack Fault Proofs

Optimism proposed Upgrade 20 on Sept. 9, 2026, to replace OP Stack output-root fault-proof games with super-root games, requiring a coordinated proof-service cutover but no L2 hardfork. The Upgrade 20 proposal targets Sepolia chains for Sept. 17 and OP Mainnet, Ink, Soneium and Unichain for Sept. 24, subject to governance approval and a seven-day testnet soak. Ordinary node, sequencer and batcher configurations do not change; operators who also run proof infrastructure have work to do.

How do Super Root Dispute Games work?

Super Root Dispute Games replace a commitment to one chain’s output at an L2 block number with a timestamped super root that can eventually hold commitments for several chains. At this stage, however, each game still contains only one chain’s output root. Upgrade 20 prepares the dispute format for interoperability without enabling shared cross-chain disputes.

The sequencer continues ordering L2 transactions, while the batcher publishes the data needed to derive the chain to Ethereum. An op-proposer reads a commitment from a Super Root RPC and creates a game through the chain’s L1 DisputeGameFactory. On permissionless chains, an op-challenger independently reconstructs the state using the Kona fault-proof program and Ethereum-posted data. A disagreement enters the existing interactive bisection process, and an accepted result advances the commitment recorded through the AnchorStateRegistry.

The main structural change is that an incorrect timestamp is tested inside the fault-proof state transition. The separate L2 block-number challenge used by output-root games disappears.

What must OP Stack operators change?

Chain operators must switch the services surrounding the fault-proof contracts at the same time as the L1 upgrade executes.

• Point op-proposer at --superroot-rpcs and select game type 5 for permissioned proofs or type 9 for permissionless proofs.
• Add a Super Root RPC to op-challenger and op-dispute-mon, while preserving old settings until outstanding output-root games finish.
• Stage and reproduce the reviewed Kona interoperability-variant prestate before enabling permissionless super-root games.
• Keep any node endpoint serving Super Root RPC requests reachable throughout the cutover.

A failed RPC cutover could stop new proposals or blind monitoring even if the L2 chain keeps producing blocks. A wrong root or timestamp can be challenged on a permissionless chain. Permissioned games take a different path: they resolve immediately for the defender and rely on an air gap plus Guardian blacklisting to stop an invalid proposal.

Does Upgrade 20 reduce fees or hardware demands?

No: Upgrade 20 changes proof commitments, not transaction execution, sequencing or data availability. It therefore offers no announced increase in throughput or reduction in user fees. The permissionless game reuses the established bond, clock, bisection and resolution machinery, so the proof workload is closer to a format migration than a new proving system.

The proposal publishes no measured comparison of challenger CPU, memory, bandwidth or L1 gas use. Operators gain an additional Super Root RPC dependency and must host the correct Kona prestate, but ordinary nodes acquire no new configuration requirement. Claims about neutral resource use remain targets until the Sepolia soak and production disputes supply measurements.

Is Upgrade 20 a practical improvement?

Yes, narrowly: moving to the timestamped format now reduces the number of dispute-system changes required when multi-chain interoperability arrives. It also lets old and new games resolve in parallel, limiting migration disruption. The contracts received a completed Spearbit/Cantina review on Aug. 21, according to the proposal.

This is operational groundwork, not a scaling release. It does not activate interoperability, consolidate dispute factories or prove that multi-chain super-root games work under production load. Upgrade 20 improves the upgrade path, but its larger payoff remains unproven until the root actually covers several chains.