Nethermind: EIP-8141 could lift its validation cap to 300k gas if nodes pause validation during block import
Three Nethermind researchers, Stefano De Angelis, Daniil Ankushin and Manuel Arto, published a benchmark on Ethereum Research on 6 November that asks how large the validation budget of EIP-8141 frame transactions can safely be. In a frame transaction the signature check is arbitrary EVM code in a VERIFY frame. A node must run that code before admitting the transaction to its mempool, and if the check fails nobody pays for the work, so the proposal caps it with MAX_VERIFY_GAS, currently 100,000 gas.
The authors argue that 100,000 is too low for important uses. A privacy transfer with a single-note Groth16 spend needs at least about 212,000 gas, hash-based post-quantum signature schemes cost between 94,000 and 276,000, a PLONK verifier about 298,000, while lattice-based post-quantum schemes in pure EVM run from about 1.2 million to 4.7 million. Raising the cap, however, lets an attacker send transactions that burn the whole budget and then fail, taking CPU away from block processing for free.
They flooded a Nethermind node with such transactions at caps from 100,000 to 500,000 on three machines, including a worst case of one CPU core, and counted a load as safe if it added no more than 25 milliseconds to importing or building a median mainnet block. A builder-class machine absorbed about 30 million gas per second of validation work within that budget; the slowest machine, standing in for a full node, managed only about 15 million. Since Nethermind allows up to about 100 failing transactions per second per connection, that limit leaves only the current 100,000 cap within budget for full nodes.
The proposed fix is a client-side protection: while the node processes a block, it starts no validation of gossiped frame transactions. With it, import delay dropped to about 2 milliseconds or less on every machine, and a 300,000 cap stayed within budget at 100 transactions per second, keeping block processing within 50 milliseconds even on large mainnet blocks. The cost moves to the mempool: under that flood the nodes admitted only 12 to 27 percent of honest frame transactions.

What it means
The study turns a design debate into numbers. A higher cap would make privacy and post-quantum accounts practical with frames, but only if clients adopt protections like Nethermind's, and the mempool trade-off means honest users could still be crowded out during an attack.