fix: disable post-Jovian DA footprint cap in benchmark

[meyer9]

Problem

In the mainnet config benchmark (Scale Base over 150M gas limit), the mainnet-transfer-only and mainnet-account-create-full-block payloads plateau at exactly 52.6% of the configured gas limit regardless of node capacity:

Payload	250M limit gas/per_block	Fill	Per-flashblock headroom
storage-create-full-block	245 M	98%	0.9–1.9%
mainnet-transfer-only	131 M	52.6%	47.4% (every block)
mainnet-account-create-full-block	131 M	52.6%	47.4% (every block)

(Run test-1779411190779930; the 47.4% headroom is identical across all 900 blocks and across all 150M/200M/250M gas limits, with a range of 0.10–0.36%.)

Root cause

The Jovian DA-footprint cap in base/crates/builder/core/src/execution.rs::is_tx_over_limits was firing before the gas cap:

let tx_da_footprint = total_da_bytes_used.saturating_mul(da_footprint_gas_scalar as u64);
if tx_da_footprint > limits.block_da_footprint_limit.unwrap_or(limits.block_gas_limit) {
    return Err(TxnExecutionError::DAFootprintLimitExceeded { .. });
}

The benchmark hardcoded DAFootprintGasScalar: 400, and block_da_footprint_limit defaults to block_gas_limit. Effective rule:

cumulative_da_bytes × 400 ≤ block_gas_limit
→ max DA bytes/block = block_gas_limit / 400

For ~100 B-per-tx transferonly:

Gas limit	Predicted tx/block (= limit / 400 / 100)	Observed tx/block
150 M	3,750	3,751
200 M	5,000	5,000
250 M	6,250	6,251

Variance < 0.03% across three limits — conclusive.

storage-create-full-block was unaffected because its txs are heavy in gas (~2.27 M) but light in DA (~200 B), so the gas cap fires first.

Fix

Set DAFootprintGasScalar to 0 in generatePayloadAttributes. The check then computes tx_da_footprint = bytes × 0 = 0 ≤ limit and always passes. The scalar is serialized into the L1 attributes deposit tx and written to L2 state; the builder reads it via fetch_da_footprint_gas_scalar per block. Setup blocks (which run with GasLimitSetup = 1e9) propagate the new scalar before benchmark blocks start.

Expected outcome on next rerun

• mainnet-transfer-only @ 250M: 131 M → ~245+ M gas/per_block
• mainnet-account-create-full-block @ 250M: was bound by both DA cap and EL speed (989 ms get_payload); will now be a true EL-limited measurement
• storage-create-full-block @ 250M: unchanged (~245 M, already gas-capped)
• base-mainnet-simulation @ 250M: still won't fill (separate worker-side scaleFactor bug, not addressed here)

Verification

• go build ./runner/... clean
• go vet ./runner/network/consensus/... clean
• No tests in the consensus package exercise this code path
• Mechanism source-verified against base/crates/builder/core/src/execution.rs

Rationale

The benchmark's question is "how much L2 gas/sec can the EL execute?". L1 DA capacity is a separate concern measured elsewhere; baking it into the gas-throughput benchmark masks the EL-throughput signal for cheap-tx workloads.

Draft because end-to-end validation requires a single ~12 h mainnet rerun.

[cb-heimdall]

🟡 Heimdall Review Status

Requirement	Status	More Info
Reviews	🟡 0/1	Denominator calculation Show calculation 1 if user is bot 0 1 if user is external 0 2 if repo is sensitive 0 From .codeflow.yml 1 Additional review requirements Show calculation Max 0 0 From CODEOWNERS 0 Global minimum 0 Max 1 1 1 if commit is unverified 0 Sum 1