Meeting Summary:
ACDT number 91 focused on DevNet 7 and 8 planning discussions held on August 10, 2026. The meeting covered several key topics including DevNet 7 status updates, where Stefan reported overall progress with some syncing issues being investigated, and Toni presented findings on a potential attack vector involving thousands of payloads causing missed slots and network congestion. The team discussed concerns about repricing numbers for DevNet 8, with Jochem explaining that final numbers cannot be confirmed until additional benchmark testing is completed, leading to a potential delay in the testnet launch. The group also addressed a proposal for a 200 million gas limit bump at the fork boundary, with debate over whether to implement this change before or after the Glasperse fork, and discussed the mandatory implementation of EIP 8070 for DevNet 8. Finally, the meeting covered circuit breaker design progress across different clients, with updates from Teku and Nimbus on their implementations, though Lighthouse representation was limited.
Click to expand detailed summary
The meeting began with participants discussing the agenda for ACDT number 91 on August 10, 2026. Mario announced that the agenda was light and focused on discussions about planning for DevNet 8. The group decided to stay in the current call for the EL breakout since there were no specific EL breakout topics. Mario then invited someone to provide updates on DevNet 7 status, though the transcript ended before this topic was addressed.
Stefan reported that overall progress was going well with no finality stalls in recent days, though there were some syncing issues with prison nodes that have been fixed. Toni explained his attack involving thousands of correctly signed payloads of 9.5 megabytes each that caused missed slots and payload propagation issues, particularly affecting Lighthouse and Prism nodes. Dustin clarified that while Nimbus wasn’t directly affected by the first-order propagation issue, it experienced delays in envelope acceptance due to high volume traffic to ELs, though no single severe bug was identified on the Nimbus side.
The team discussed a security issue involving 95,000 transactions that could potentially bypass network checks due to low gas limits. Toni explained that while the CL should fail on early validation due to the high transaction volume exceeding current gas limits, attackers could still fit many transactions within the existing 60 million gas limit. Terence raised questions about potential bottlenecks on the CL side, particularly regarding state transition functions and hash verification, while Toni agreed to test the issue on DefNet 8 and coordinate with the team to capture performance profiles during testing.
The team discussed an issue with re-propagation of signed execution payload envelopes, where Toni identified that some clients were re-propagating envelopes they had already seen, potentially causing bottlenecks. The group agreed to test this issue on DevNet 8 rather than DevNet 7, as it appears to be a specification implementation gap rather than a bug. The discussion also touched on an EELS release 8.1.0 that matches repricing numbers in the existing EIPs repo and adds test coverage, though there were concerns that the numbers might change again.
The team discussed concerns about the finality of repricing numbers provided last week, with Jochem-Brouwer explaining they cannot confirm these are the final numbers due to ongoing testing on different infrastructure. The team plans to run benchmarks on the original domain infrastructure to verify if the numbers remain consistent, with two possible outcomes: either the numbers will be confirmed as final or changes will be needed. The discussion included clarification about whether Devnet 8 would be public, with Stefan confirming it is currently planned to be public.
The team discussed concerns about the accuracy of proposed numbers for a Thursday deployment and decided to delay until after the ACDE meeting where final numbers would be confirmed. Jochem explained they were not comfortable signing off on current numbers and wanted to ensure accuracy rather than making multiple adjustments. The group agreed to wait for the ACDE presentation to get final numbers before proceeding with the next ELs release and testing phase.
The team discussed using a 70MB Genesis SSZ state for DevNet 8 to test encoding issues that occurred in DevNet 7. While some clients expressed concerns about the size, the group agreed it was acceptable since the state would be dynamically downloaded rather than included in binaries. The discussion then shifted to a proposal by Barnabas to implement a 200MB gas limit bump at the fork boundary to accommodate the new repricing values and prevent transactions from failing due to insufficient gas limits.
The team discussed implementing a new gas limit feature for the GLAMsterdam fork, with Barnabas proposing to set the gas limit to 200 million through client releases. Several participants, including Dustin and Nico, expressed concerns about creating a dual gas limit system that could be dangerous if activated before the intended fork boundary. The group agreed to further discuss and draft a proper proposal for implementing this feature without requiring two separate releases. Additionally, the team approved making EIP 8070 mandatory for DevNet 8, and Parithosh provided updates on circuit breaker implementations across different clients, with Mehdi confirming that Teku’s implementation is already in the DevNet 8 branch and needs stricter configuration values to trigger properly.
Next Steps:
- Toni: Reproduce the payload attack on DevNet 8 and communicate the start time to all clients for data collection.
- Toni: Produce a write-up for the payload attack issue.
- Jochem-brouwer: Present the final repricing numbers at ACDE and confirm if they are final or need changes.
- Stefan: Proceed with a 70MB Genesis state for DevNet 8 to test the SSC encoding issue.
- Barnabas: Draft a new proposal for the 200M gas limit bump at the Glamos fork, considering the discussion and avoiding two releases.
- Mehdi: Provide the flag values to enable and trigger the Teku circuit breaker on DevNet 8.
- Lighthouse team: Provide an update on their circuit breaker implementation for DevNet 8.
- Nimbus team: Implement the circuit breaker for payloads, similar to their current design for blocks, for DevNet 8.
- All clients with Snap 2 and cell-level deltas implementation: Ensure it is rebased on the DevNet 8 branch for testing.
- Jochem-brouwer: Follow up with the EIP editor to merge the updates for EIP 1202.
Recording Access: