EIP-8131 Empirical Impact

Isolated EIP-8131 impact, assuming EIP-7976 is active. Mainnet, last 6 months (2025-11-08 to 2026-05-08), via Xatu.

   
All mainnet txs 376,262,548
All mainnet senders 54,160,707
Type-4 txs 2,658,513 (0.71% of all)
Type-4 senders 208,978 (0.39% of all)
Affected txs 3,288 (0.124% of type-4 / 0.001% of all)
Affected senders 689 (0.330% of type-4 senders)
Total extra gas (6 mo) 27,648,764
Δ per affected tx (median / p95 / p99) 6,464 / 19,392 / 71,104
Mean / median Δ as share of gas_C_7976 (EIP-7976 baseline) 5.53% / 5.18%
Affected-tx success rate 22.0% (725 succ / 2,563 reverted; vs. 93.6% across all type-4)

Median Δ = FLOOR_COST_PER_AUTH exactly; 80.8% of affected txs are 1-auth delegations whose intrinsic + execution already sat below floor_C, so EIP-8131 layers cleanly on top. The bulk of the floor-bound population are reverted txs whose execution path doesn’t burn enough gas to clear the new floor.

Top affected senders

# sender affected txs extra gas max auths total auths
1 0x59aab1bd…25e16b 296 7,757,005 31 2,138
2 0xfc152f3c…40e2b 7 1,639,149 130 640
3 0xe46f81fa…0b99ae 120 771,627 1 120
4 0xb9d00655…24da7c 99 610,031 1 99
5 0x35f8f66c…b4586f 98 597,045 1 98
6 0xbb4d1438…41bcf 90 551,606 1 90
7 0xec103c6e…58b915 53 338,304 1 53
8 0x34f41c98…35404f 53 332,248 1 53
9 0x6d2ccc86…97df85 50 312,367 1 50
10 0x7d8f6522…1af30a 50 312,316 1 50

The top sender alone accounts for ~28% of all extra gas. #2 is the cleanest “bypass-shaped” sender: 7 txs at up to 130 auths each.

Affected by auth count

auths type-4 affected extra gas share
1 2,138,203 2,994 17,493,597 63%
2 75,539 76 800,509 3%
3 63,608 54 876,717 3%
4-5 93,768 43 996,673 4%
6-10 118,930 29 1,042,062 4%
11-20 98,125 48 3,085,595 11%
21-50 57,085 37 1,714,462 6%
51-200 10,949 7 1,639,149 6%
>200 2,306 0 0 0%

Method

num_authorizations estimated from tx.size − tx.call_data_size (≈92 bytes/auth + 117-byte envelope on mainnet); bucket modes verified against the empirical histogram (boundary buckets disagree by ≤1 auth). Today’s gas_used used as proxy for intrinsic + execution, exact for 99.95% of type-4 txs (1,248 rows / 0.05% have gas_used == floor_A_today, where the proxy is loose; the isolated metric Δ = floor_D − floor_C is invariant to the proxy in that regime, so unaffected). 0 spec violations (gas_used < floor_A_today). Long auth-tail (≥51 auths) contributes <6% of total Δ.

Appendix: context deltas vs. today

scenario affected senders total extra gas median Δ mean Δ / g
EIP-8131 only (10/token floor) 1,595 358 10,670,600 6,464 6.2%
EIP-7976 only 2,657 565 175,924,489 31,596 32.3%
EIP-7976 + EIP-8131 3,288 689 203,573,253 33,062 32.9%

Isolated EIP-8131 portion = 27,648,764 (target row above).