EIP-1884 and EIP-2929 are good references — but they worked in a fundamentally different era. When EIP-1884 shipped (December 2019), the base fee was 5+ gwei. When EIP-2929 shipped (April 2021), it was 100+ gwei. Repricing gas units mattered because each unit cost real money.
That era is over. The base fee now sits between 0.09 and 0.6 gwei and it’s only going lower — every capacity upgrade (Glamsterdam targeting 100-300M gas limits) pushes it further toward zero. EIP-1559’s base fee only rises above 50% utilization, and the entire scaling roadmap ensures it doesn’t. EIP-8037 reprices a new account to 131,488 gas at 100M gas limit — at 0.1 gwei, that’s $0.027. Maria Silva’s own EIP-8037 optimization assumes a 1 gwei baseline — without a floor enforcing that, her model’s state growth outcomes don’t hold.
You say “the benefits gained from reducing spam do not seem sufficient to offset the cost.” An EIP-1559 floor isn’t just about spam — it’s about state growth, the burn, and the integrity of the fee market itself. Vitalik’s recent post on hyper-scaling state identifies state as the hardest scaling bottleneck with no magic bullet. At near-zero base fees, this externality is completely unpriced. And EIP-1559’s three original promises — fee predictability, auction efficiency, and the burn — all break at zero. Fees oscillate between near-zero and sudden spikes instead of smoothing. Inclusion priority shifts to tips, recreating the first-price auction EIP-1559 replaced. And the burn is dead — ultrasound.money’s 30-day data shows 8K ETH/year burn against 1,001K issuance, less than 1%.
The cost of a floor? A transfer goes from $0.004 to $0.04. The cost of not having one is a broken fee market, unpriced state growth, permanent inflation, and validators drifting toward MEV dependence.