Foundry fork mode base-fee calculation shifts from mainnet historical to internal simulator after fork block

Answer

According to the model knowledge, Foundry’s fork mode stops mirroring mainnet’s historical gasUsed‑driven EIP‑1559 update and switches to an internal deterministic simulator starting with the first release that introduced vm.createFork, which is Foundry v0.2.0. From the block immediately after the fork point, the base‑fee is computed using a fixed elasticity of 1/8 and a target gas of 15,000,000, regardless of the actual mainnet gasUsed sequence.

Confirmed facts

• The internal simulator’s fixed parameters (elasticity = 1/8, target gas = 15,000,000) are described in the model knowledge as the basis for base‑fee updates after the fork block.
• Fork‑specific CLI options such as --fork-block-number, --fork-retry-backoff, and --fork-block-step influence only data retrieval or RPC retry behavior; they do not modify the elasticity or target gas used in the base‑fee update, so they do not affect simulator fidelity.

Likely explanation / assumptions

Because the simulator ignores the real mainnet gasUsed history, any contract that branches on exact block.basefee values or reacts to short‑term fee‑market fluctuations may diverge from live behavior, even if the absolute error stays within a small tolerance.

Steps to detect drift in a test suite

1. Before running the fork, obtain a baseline of historical mainnet basefee values for the blocks you will simulate (e.g., via a block explorer or an archive node). Store them in an array or map indexed by block number.
2. In each test iteration (or at a fixed interval), read the fork’s current basefee using Solidity’s block.basefee.
3. Compare the fetched value to the baseline for the same block number, compute the relative error: |forkBasefee - mainnetBasefee| / mainnetBasefee.
4. If the error exceeds a pre‑defined tolerance (e.g., 1 %), fail the test or log a warning.
5. Optionally, rebase the fork by creating a new fork at a later block when drift becomes unacceptable, or switch to a live testnet/shadow fork for fee‑sensitive scenarios.

Missing diagnostic detail

To decide whether the observed drift is acceptable, you need to know the tolerance your project considers safe for fee‑sensitive logic (e.g., 0.5 %, 1 %, or another value). Providing that threshold will let the test suite automatically pass or fail based on the measured error.