Improve Reliability of the CCIP Gateway

[alainncls]

Description:

The CCIP Gateway currently lacks sufficient reliability in handling RPC failures, which can lead to downtime a when a single RPC provider (e.g., Infura) is unavailable. To address this issue, implement one of the following solutions:

1. Implement a fallback RPC URL at the gateway level:

   • Add support for multiple RPC endpoints directly in the gateway.
   • Automatically switch to a secondary RPC provider in case of a failure.

2. Deploy a second gateway instance (recommended):

   • Use a separate RPC provider for the second gateway instance (e.g., the Linea internal RPC node).
   • Deploy a new LineaSparseProofVerifier referencing the 2 URLs.
   • Update the L1Resolver contract to reference the new LineaSparseProofVerifier address.

Acceptance Criteria:

• The CCIP Gateway should remain operational during an RPC provider failure.
• The chosen solution must minimize impact on the current infrastructure while enhancing resilience.
• For solution 2, proper redeployment of required contracts (e.g., L1Resolver) must be included.

Note:

Solution 2 is the recommended approach as it aligns with CCIP-Read best practices and ensures greater reliability.

[alainncls]

Solution 1:

• Small code modification
• Potential additional cost for RPC provider?
• Same devops chart with additional RPC config
• No contract to deploy
  => Easy to implement & support

Solution 2:

• No code modification, only configuration (DNS + deploy script)
• Recommended by the CCIP standard
• Requires 2 devops charts + 2 autoscaling instances
• 2 contracts to redeploy
• Requires Security Council sign off
• Potential additional cost for RPC provider?
  => Easy to implement

Conclusion:

• Solution 1 is lightweight and less expensive than solution 2
• Solution 2 is ideal but more expensive and requires more devops config/work
• Our recommendation: Solution 1 because practically speaking, only the 1st gateway will be called 99% of the time (so the 2nd instance is idle).

[alainncls]

ENS L1 makes the call to the contract, it reverts, but the interpretation of this revert is handled by ENS L1, not in our code.
=> Having 2 gateways doesn't have any impact on our UI