Understanding L2 Architecture Types

Before comparing specific networks, understand the four fundamental L2 architecture categories and their security trade-offs.

1. Optimistic Rollups

Optimistic rollups execute transactions off-chain, post compressed transaction data to Ethereum L1, and assume transactions are valid unless challenged during a dispute period (typically 7 days).

How it works:

1. •Users submit transactions to the L2 sequencer
2. •Sequencer batches transactions and posts compressed calldata to L1
3. •A state root is published to L1
4. •During the challenge window (7 days), anyone can submit a fraud proof
5. •If a fraud proof succeeds, the invalid state transition is reverted

Security model: Inherits Ethereum security with the caveat that at least one honest verifier must be watching and capable of submitting fraud proofs within the challenge period. In practice, centralized sequencers are a single point of failure for liveness (not safety).

Key implementations: Arbitrum One, Optimism, Base, Mantle, Blast

2. ZK Rollups (Validity Rollups)

ZK rollups execute transactions off-chain, generate cryptographic validity proofs (ZK-SNARKs or ZK-STARKs) proving the correctness of every state transition, and verify these proofs on L1.

How it works:

1. •Users submit transactions to the L2 sequencer
2. •Sequencer executes transactions and generates a validity proof
3. •Proof + compressed state diff posted to L1
4. •L1 verifier contract validates the proof (instant finality)
5. •No dispute period needed — mathematically verified

Security model: Strongest L2 security — validity proofs cryptographically guarantee correct execution. Compromising a ZK rollup requires breaking the underlying cryptographic assumptions, which is considered computationally infeasible.

Key implementations: zkSync Era, Starknet, Polygon zkEVM, Scroll, Linea, Taiko

3. Sidechains

Sidechains are independent blockchains with their own consensus mechanisms, connected to Ethereum via a bridge. They do NOT inherit Ethereum's security.

Security model: Relies on the sidechain's own validator set. If validators collude or the consensus mechanism is compromised, user funds can be stolen. This is fundamentally weaker than rollups.

Key implementations: Polygon PoS (transitioning to zkEVM), Ronin, Gnosis Chain

4. Validiums

Validiums generate ZK validity proofs (like ZK rollups) but store transaction data off-chain via a Data Availability Committee (DAC) instead of on Ethereum L1. This dramatically reduces costs but introduces a data availability trust assumption.

Security model: Proofs are valid, but if the DAC withholds data, users cannot reconstruct state or prove asset ownership. Hybrid models (Volition) let users choose between L1 data availability (rollup mode) and off-chain DA (validium mode) per transaction.

Key implementations: Immutable X, StarkEx-based apps (dYdX v3), ApeChain

Architecture Comparison Matrix

Head-to-Head: Top L2 Networks Compared

Real-World Performance Data (March 2026 averages)

Fee Deep-Dive: Real Transaction Costs

Post-EIP-4844 (Dencun upgrade, March 2024), L2 fees dropped 90-99% thanks to blob data availability. Here is what actual operations cost on each L2 (March 2026):

Base offers the lowest fees among major L2s due to Coinbase's sequencer optimization and aggressive blob data usage. Starknet achieves very low costs through its STARK proof batching, which amortizes proof costs across many transactions.

Cost structure breakdown: L2 fees consist of two components:

1. •L2 execution fee — Gas cost for computation on the L2 itself (negligible on most L2s)
2. •L1 data fee — Cost to post transaction data to Ethereum L1 (dominant cost component)

With EIP-4844 blobs, the L1 data fee dropped from $0.50-5.00 per transaction to $0.001-0.05, making L2s viable for mass-market applications.

Optimistic Rollups: Detailed Comparison

Arbitrum One vs. Optimism vs. Base

Arbitrum One is the conservative choice: most battle-tested, largest DeFi TVL, most protocol integrations. Its Nitro upgrade (custom WASM-based VM) provides higher throughput than standard OP Stack chains.

Optimism is building the Superchain — a network of OP Stack chains that share a security model and bridging infrastructure. Member chains include Base, Worldchain, Zora, Mode, and 20+ others. If network effects matter to your strategy, the Superchain is the most ambitious horizontal scaling play.

Base is the distribution play. Coinbase's 100M+ verified users create unmatched onboarding potential. Coinbase Smart Wallet (using ERC-4337 account abstraction) lets users transact on Base without knowing they are using a blockchain. For consumer apps, this distribution advantage is difficult to replicate.

Arbitrum Orbit vs. OP Stack: Building Your Own L2/L3

ZK Rollups: Detailed Comparison

zkSync Era vs. Starknet vs. Polygon zkEVM

EVM Compatibility Types (Vitalik's taxonomy):

For most builders: Type 2 (Polygon zkEVM, Scroll) offers the easiest migration from Ethereum L1, since existing Solidity contracts deploy with zero or minimal modifications. Type 4 (zkSync, Starknet) offers performance advantages but may require code modifications and specialized tooling.

ZK-Provers: Performance and Cost

The ZK prover is the most computationally expensive component of a ZK rollup. Real-world prover metrics:

Proof generation costs are declining rapidly due to hardware acceleration (FPGA/ASIC provers by Cysic, Ulvetanna, Ingonyama) and proof aggregation. By late 2026, proof costs are projected to drop 60-80% from current levels.

Choosing the Right L2: Decision Framework

By Use Case

By Priority

Deployment Guide: Launching on an L2

Step 1: Contract Deployment Differences

For most L2s, deployment is nearly identical to Ethereum L1. Key differences:

Optimistic rollups (Arbitrum, Base, Optimism):

• •Hardhat/Foundry config: just change the RPC URL and chain ID
• •99.9% Solidity compatibility — almost all L1 contracts deploy without changes
• •Differences: block.number returns L2 block number (not L1), tx.gasprice includes L1 data fee

ZK rollups (zkSync):

• •Use hardhat-zksync plugin or zksync-cli
• •Contracts compile via zksolc (modified Solidity compiler)
• •CREATE2 and CREATE behave differently (address derivation includes bytecode hash)
• •Some precompiles may not be supported or have different gas costs

Starknet:

• •Contracts written in Cairo (Rust-like language), not Solidity
• •Different account model (native account abstraction)
• •Requires learning Starknet-specific tooling (Scarb, Starkli)

Step 2: Bridge Integration

Every L2 needs a bridge for asset transfers from L1. Options:

For user-facing bridges, liquidity network bridges (Across Protocol, Stargate) provide the best balance of speed and security. For programmatic bridge integration, canonical bridges are safest but slowest.

Step 3: Sequencer and MEV Considerations

All major L2s currently run centralized sequencers, creating several important considerations:

Shared sequencer roadmap: Projects like Espresso Systems, Astria, and Radius are building shared sequencers that would decentralize transaction ordering across multiple L2s. This is expected to go live on select L2s in late 2026 to early 2027, significantly reducing centralization concerns.

Step 4: Data Availability Options

With EIP-4844, L2s post data as "blobs" to Ethereum L1. For app-chains needing even lower costs, alternative DA layers are available:

For most builders: Ethereum blobs provide the strongest security guarantees and are sufficient for standard rollups. Alternative DA layers make sense for high-throughput app-chains (gaming, social) where data costs would otherwise be prohibitive.

The Rollup Stack Wars: OP Stack vs. ZK Stack vs. Polygon CDK vs. Arbitrum Orbit

The battle for L2 proliferation is defined by these four frameworks:

The Superchain thesis (Optimism): All OP Stack chains share the same bridge contracts, enabling atomic cross-chain messages. As of March 2026, 30+ chains are part of the Superchain, including Base (Coinbase), Worldchain (Worldcoin), Zora, Mode, Cyber, and Redstone. This creates the largest L2 network effect.

The Hyperchain thesis (zkSync): ZK Stack chains can share proofs and settle via a single proof on L1, theoretically enabling the cheapest interoperability. Earlier stage than the Superchain but architecturally compelling.

The AggLayer thesis (Polygon): Polygon CDK chains share a unified proof aggregation layer, enabling cross-chain state verification. Partners include Immutable (gaming), Astar (Japan ecosystem), and OKX's X Layer.

Interoperability: The Missing Piece

Cross-L2 interoperability remains the biggest unsolved challenge. Moving assets between Arbitrum and Base today requires a bridge transaction (1-5 minutes, $0.50-5.00 in fees and slippage). The vision: seamless asset movement across all L2s without explicit bridging.

Solutions in Development

For builders deploying across multiple L2s today, intent-based bridges (Across, UniswapX cross-chain) provide the best user experience. For long-term architecture, aligning with one of the major stack ecosystems (Superchain, Hyperchains, AggLayer) will provide native interoperability as these systems mature.

Security Maturity: L2Beat Risk Assessment

L2Beat tracks the security maturity of every L2 across five risk dimensions. As of March 2026:

Key takeaway: Arbitrum One has the most decentralized security model with live fraud proofs (BoLD), open proposing, and a 7-day exit window. Most ZK rollups still have permissioned proposers and longer upgrade timelocks. For high-value applications, check L2Beat's risk assessment before choosing your L2.

For L2 deployment support, browse development partners in our directory who can help with smart contract migration, bridge integration, and cross-chain architecture. For security audits of your L2 deployment, see our security partners. For broader ecosystem research, check The Signal intelligence reports.

Frequently Asked Questions

What is the cheapest Layer 2 for deploying a dApp?

Base offers the lowest transaction fees among major L2s, with simple transfers costing $0.001-0.005 and complex DeFi transactions under $0.01. Starknet is competitive with similarly low fees. For app-chains, using alternative data availability layers (Celestia, EigenDA) can reduce costs further. Monthly operating costs for a moderate-traffic dApp on Base range from $100-500 in sequencer fees.

What is the difference between optimistic and ZK rollups?

Optimistic rollups assume transactions are valid and use a 7-day challenge period where fraud proofs can dispute incorrect state transitions. ZK rollups generate mathematical proofs (ZK-SNARKs/STARKs) that verify every transaction is correct, enabling faster L1 finality (minutes to hours instead of 7 days). ZK rollups offer stronger security guarantees but are more computationally expensive to produce proofs and may have limited EVM compatibility.

Can I deploy existing Ethereum smart contracts on Layer 2 without changes?

On optimistic rollups (Arbitrum, Optimism, Base): yes, with 99.9% compatibility. Just change the RPC URL. On Type 2 ZK rollups (Polygon zkEVM, Scroll): yes, with 99%+ compatibility and minor edge cases. On Type 4 ZK rollups (zkSync Era): Solidity code compiles but some patterns may behave differently (CREATE2 address derivation, gas costs). On Starknet: no, contracts must be rewritten in Cairo.

How do I choose between Arbitrum, Optimism, and Base?

Arbitrum for deepest DeFi liquidity and most battle-tested infrastructure. Optimism for Superchain ecosystem participation and retroactive public goods funding. Base for Coinbase's 100M+ user distribution and lowest fees. If your dApp serves DeFi power users, choose Arbitrum. If your dApp targets mainstream consumers, choose Base. If you plan to launch your own L2, evaluate OP Stack membership in the Superchain.

Is it possible to build a custom Layer 2 chain?

Yes. The four main frameworks are: OP Stack (optimistic, MIT licensed, Superchain ecosystem), Arbitrum Orbit (optimistic, free for Arbitrum settlement), ZK Stack (ZK proofs, MIT licensed), and Polygon CDK (ZK proofs, AggLayer ecosystem). Deployment takes 1-4 weeks. Operating costs range from $2,000-15,000/month depending on transaction volume and data availability choice.

What happens to my funds if a Layer 2 sequencer goes offline?

All major L2s have a "force inclusion" mechanism that allows users to submit transactions directly to L1, bypassing the sequencer. On Arbitrum, you can use the Delayed Inbox contract. On Optimism/Base, the OptimismPortal contract. This means your funds are never permanently stuck, but the experience during sequencer downtime is degraded (slower transactions, higher costs).

Will Layer 2 fees increase as Ethereum blob space fills up?

Potentially, but this is being addressed. Ethereum's PeerDAS upgrade (expected late 2026) will increase blob capacity from 6 per block to 64-128, a 10-20x increase. Additionally, L2s are developing proof aggregation and data compression techniques that reduce per-transaction data requirements. For the foreseeable future, L2 fees will remain orders of magnitude cheaper than L1.

What is the Superchain and why does it matter?

The Superchain is Optimism's vision of a network of interoperable OP Stack chains that share security, bridging, and sequencer infrastructure. As of March 2026, 30+ chains participate including Base, Worldchain, and Zora. The Superchain matters because it enables atomic cross-chain transactions, shared liquidity, and a unified user experience across all member chains — solving the L2 fragmentation problem.

Understanding L2 Architecture Types

Before comparing specific networks, understand the four fundamental L2 architecture categories and their security trade-offs.

1. Optimistic Rollups

Optimistic rollups execute transactions off-chain, post compressed transaction data to Ethereum L1, and assume transactions are valid unless challenged during a dispute period (typically 7 days).

How it works:

1. •Users submit transactions to the L2 sequencer
2. •Sequencer batches transactions and posts compressed calldata to L1
3. •A state root is published to L1
4. •During the challenge window (7 days), anyone can submit a fraud proof
5. •If a fraud proof succeeds, the invalid state transition is reverted

Security model: Inherits Ethereum security with the caveat that at least one honest verifier must be watching and capable of submitting fraud proofs within the challenge period. In practice, centralized sequencers are a single point of failure for liveness (not safety).

Key implementations: Arbitrum One, Optimism, Base, Mantle, Blast

2. ZK Rollups (Validity Rollups)

ZK rollups execute transactions off-chain, generate cryptographic validity proofs (ZK-SNARKs or ZK-STARKs) proving the correctness of every state transition, and verify these proofs on L1.

How it works:

1. •Users submit transactions to the L2 sequencer
2. •Sequencer executes transactions and generates a validity proof
3. •Proof + compressed state diff posted to L1
4. •L1 verifier contract validates the proof (instant finality)
5. •No dispute period needed — mathematically verified

Security model: Strongest L2 security — validity proofs cryptographically guarantee correct execution. Compromising a ZK rollup requires breaking the underlying cryptographic assumptions, which is considered computationally infeasible.

Key implementations: zkSync Era, Starknet, Polygon zkEVM, Scroll, Linea, Taiko

3. Sidechains

Sidechains are independent blockchains with their own consensus mechanisms, connected to Ethereum via a bridge. They do NOT inherit Ethereum's security.

Security model: Relies on the sidechain's own validator set. If validators collude or the consensus mechanism is compromised, user funds can be stolen. This is fundamentally weaker than rollups.

Key implementations: Polygon PoS (transitioning to zkEVM), Ronin, Gnosis Chain

4. Validiums

Validiums generate ZK validity proofs (like ZK rollups) but store transaction data off-chain via a Data Availability Committee (DAC) instead of on Ethereum L1. This dramatically reduces costs but introduces a data availability trust assumption.

Security model: Proofs are valid, but if the DAC withholds data, users cannot reconstruct state or prove asset ownership. Hybrid models (Volition) let users choose between L1 data availability (rollup mode) and off-chain DA (validium mode) per transaction.

Key implementations: Immutable X, StarkEx-based apps (dYdX v3), ApeChain

Architecture Comparison Matrix

Head-to-Head: Top L2 Networks Compared

Real-World Performance Data (March 2026 averages)

Fee Deep-Dive: Real Transaction Costs

Post-EIP-4844 (Dencun upgrade, March 2024), L2 fees dropped 90-99% thanks to blob data availability. Here is what actual operations cost on each L2 (March 2026):

Base offers the lowest fees among major L2s due to Coinbase's sequencer optimization and aggressive blob data usage. Starknet achieves very low costs through its STARK proof batching, which amortizes proof costs across many transactions.

Cost structure breakdown: L2 fees consist of two components:

1. •L2 execution fee — Gas cost for computation on the L2 itself (negligible on most L2s)
2. •L1 data fee — Cost to post transaction data to Ethereum L1 (dominant cost component)

With EIP-4844 blobs, the L1 data fee dropped from $0.50-5.00 per transaction to $0.001-0.05, making L2s viable for mass-market applications.

Optimistic Rollups: Detailed Comparison

Arbitrum One vs. Optimism vs. Base

Arbitrum One is the conservative choice: most battle-tested, largest DeFi TVL, most protocol integrations. Its Nitro upgrade (custom WASM-based VM) provides higher throughput than standard OP Stack chains.

Optimism is building the Superchain — a network of OP Stack chains that share a security model and bridging infrastructure. Member chains include Base, Worldchain, Zora, Mode, and 20+ others. If network effects matter to your strategy, the Superchain is the most ambitious horizontal scaling play.

Base is the distribution play. Coinbase's 100M+ verified users create unmatched onboarding potential. Coinbase Smart Wallet (using ERC-4337 account abstraction) lets users transact on Base without knowing they are using a blockchain. For consumer apps, this distribution advantage is difficult to replicate.

Arbitrum Orbit vs. OP Stack: Building Your Own L2/L3

ZK Rollups: Detailed Comparison

zkSync Era vs. Starknet vs. Polygon zkEVM

EVM Compatibility Types (Vitalik's taxonomy):

For most builders: Type 2 (Polygon zkEVM, Scroll) offers the easiest migration from Ethereum L1, since existing Solidity contracts deploy with zero or minimal modifications. Type 4 (zkSync, Starknet) offers performance advantages but may require code modifications and specialized tooling.

ZK-Provers: Performance and Cost

The ZK prover is the most computationally expensive component of a ZK rollup. Real-world prover metrics:

Proof generation costs are declining rapidly due to hardware acceleration (FPGA/ASIC provers by Cysic, Ulvetanna, Ingonyama) and proof aggregation. By late 2026, proof costs are projected to drop 60-80% from current levels.

Choosing the Right L2: Decision Framework

By Use Case

By Priority

Deployment Guide: Launching on an L2

Step 1: Contract Deployment Differences

For most L2s, deployment is nearly identical to Ethereum L1. Key differences:

Optimistic rollups (Arbitrum, Base, Optimism):

• •Hardhat/Foundry config: just change the RPC URL and chain ID
• •99.9% Solidity compatibility — almost all L1 contracts deploy without changes
• •Differences: block.number returns L2 block number (not L1), tx.gasprice includes L1 data fee

ZK rollups (zkSync):

• •Use hardhat-zksync plugin or zksync-cli
• •Contracts compile via zksolc (modified Solidity compiler)
• •CREATE2 and CREATE behave differently (address derivation includes bytecode hash)
• •Some precompiles may not be supported or have different gas costs

Starknet:

• •Contracts written in Cairo (Rust-like language), not Solidity
• •Different account model (native account abstraction)
• •Requires learning Starknet-specific tooling (Scarb, Starkli)

Step 2: Bridge Integration

Every L2 needs a bridge for asset transfers from L1. Options:

For user-facing bridges, liquidity network bridges (Across Protocol, Stargate) provide the best balance of speed and security. For programmatic bridge integration, canonical bridges are safest but slowest.

Step 3: Sequencer and MEV Considerations

All major L2s currently run centralized sequencers, creating several important considerations:

Shared sequencer roadmap: Projects like Espresso Systems, Astria, and Radius are building shared sequencers that would decentralize transaction ordering across multiple L2s. This is expected to go live on select L2s in late 2026 to early 2027, significantly reducing centralization concerns.

Step 4: Data Availability Options

With EIP-4844, L2s post data as "blobs" to Ethereum L1. For app-chains needing even lower costs, alternative DA layers are available:

For most builders: Ethereum blobs provide the strongest security guarantees and are sufficient for standard rollups. Alternative DA layers make sense for high-throughput app-chains (gaming, social) where data costs would otherwise be prohibitive.

The Rollup Stack Wars: OP Stack vs. ZK Stack vs. Polygon CDK vs. Arbitrum Orbit

The battle for L2 proliferation is defined by these four frameworks:

The Superchain thesis (Optimism): All OP Stack chains share the same bridge contracts, enabling atomic cross-chain messages. As of March 2026, 30+ chains are part of the Superchain, including Base (Coinbase), Worldchain (Worldcoin), Zora, Mode, Cyber, and Redstone. This creates the largest L2 network effect.

The Hyperchain thesis (zkSync): ZK Stack chains can share proofs and settle via a single proof on L1, theoretically enabling the cheapest interoperability. Earlier stage than the Superchain but architecturally compelling.

The AggLayer thesis (Polygon): Polygon CDK chains share a unified proof aggregation layer, enabling cross-chain state verification. Partners include Immutable (gaming), Astar (Japan ecosystem), and OKX's X Layer.

Interoperability: The Missing Piece

Cross-L2 interoperability remains the biggest unsolved challenge. Moving assets between Arbitrum and Base today requires a bridge transaction (1-5 minutes, $0.50-5.00 in fees and slippage). The vision: seamless asset movement across all L2s without explicit bridging.

Solutions in Development

For builders deploying across multiple L2s today, intent-based bridges (Across, UniswapX cross-chain) provide the best user experience. For long-term architecture, aligning with one of the major stack ecosystems (Superchain, Hyperchains, AggLayer) will provide native interoperability as these systems mature.

Security Maturity: L2Beat Risk Assessment

L2Beat tracks the security maturity of every L2 across five risk dimensions. As of March 2026:

Key takeaway: Arbitrum One has the most decentralized security model with live fraud proofs (BoLD), open proposing, and a 7-day exit window. Most ZK rollups still have permissioned proposers and longer upgrade timelocks. For high-value applications, check L2Beat's risk assessment before choosing your L2.

For L2 deployment support, browse development partners in our directory who can help with smart contract migration, bridge integration, and cross-chain architecture. For security audits of your L2 deployment, see our security partners. For broader ecosystem research, check The Signal intelligence reports.

Frequently Asked Questions

What is the cheapest Layer 2 for deploying a dApp?

Base offers the lowest transaction fees among major L2s, with simple transfers costing $0.001-0.005 and complex DeFi transactions under $0.01. Starknet is competitive with similarly low fees. For app-chains, using alternative data availability layers (Celestia, EigenDA) can reduce costs further. Monthly operating costs for a moderate-traffic dApp on Base range from $100-500 in sequencer fees.

What is the difference between optimistic and ZK rollups?

Optimistic rollups assume transactions are valid and use a 7-day challenge period where fraud proofs can dispute incorrect state transitions. ZK rollups generate mathematical proofs (ZK-SNARKs/STARKs) that verify every transaction is correct, enabling faster L1 finality (minutes to hours instead of 7 days). ZK rollups offer stronger security guarantees but are more computationally expensive to produce proofs and may have limited EVM compatibility.

Can I deploy existing Ethereum smart contracts on Layer 2 without changes?

On optimistic rollups (Arbitrum, Optimism, Base): yes, with 99.9% compatibility. Just change the RPC URL. On Type 2 ZK rollups (Polygon zkEVM, Scroll): yes, with 99%+ compatibility and minor edge cases. On Type 4 ZK rollups (zkSync Era): Solidity code compiles but some patterns may behave differently (CREATE2 address derivation, gas costs). On Starknet: no, contracts must be rewritten in Cairo.

How do I choose between Arbitrum, Optimism, and Base?

Arbitrum for deepest DeFi liquidity and most battle-tested infrastructure. Optimism for Superchain ecosystem participation and retroactive public goods funding. Base for Coinbase's 100M+ user distribution and lowest fees. If your dApp serves DeFi power users, choose Arbitrum. If your dApp targets mainstream consumers, choose Base. If you plan to launch your own L2, evaluate OP Stack membership in the Superchain.

Is it possible to build a custom Layer 2 chain?

Yes. The four main frameworks are: OP Stack (optimistic, MIT licensed, Superchain ecosystem), Arbitrum Orbit (optimistic, free for Arbitrum settlement), ZK Stack (ZK proofs, MIT licensed), and Polygon CDK (ZK proofs, AggLayer ecosystem). Deployment takes 1-4 weeks. Operating costs range from $2,000-15,000/month depending on transaction volume and data availability choice.

What happens to my funds if a Layer 2 sequencer goes offline?

All major L2s have a "force inclusion" mechanism that allows users to submit transactions directly to L1, bypassing the sequencer. On Arbitrum, you can use the Delayed Inbox contract. On Optimism/Base, the OptimismPortal contract. This means your funds are never permanently stuck, but the experience during sequencer downtime is degraded (slower transactions, higher costs).

Will Layer 2 fees increase as Ethereum blob space fills up?

Potentially, but this is being addressed. Ethereum's PeerDAS upgrade (expected late 2026) will increase blob capacity from 6 per block to 64-128, a 10-20x increase. Additionally, L2s are developing proof aggregation and data compression techniques that reduce per-transaction data requirements. For the foreseeable future, L2 fees will remain orders of magnitude cheaper than L1.

What is the Superchain and why does it matter?

The Superchain is Optimism's vision of a network of interoperable OP Stack chains that share security, bridging, and sequencer infrastructure. As of March 2026, 30+ chains participate including Base, Worldchain, and Zora. The Superchain matters because it enables atomic cross-chain transactions, shared liquidity, and a unified user experience across all member chains — solving the L2 fragmentation problem.