Curve wants Glamsterdam judged under load, where a leveraged borrow burns 15 times a swap's gas

A leveraged borrow at 1.84 million gas against a stablecoin swap at 125,000 gives a ratio of about 14.7 ^[16]. A fee is gas used multiplied by the gas price. So whatever price a congested block clears at, the borrower pays roughly 14.7 times what the swapper pays ^[17]. In a sharp move, according to Curve, borrowers, liquidators and oracle updates all compete for the same blockspace ^[10].

Block size changes the count. At roughly 60 million gas, a block holds about 32 such borrows or 480 swaps ^[18]. At the 200 million test setting it holds about 108 borrows or 1,600 swaps, 3.3 times the room ^{[19] [20]}. That setting is Sepolia's, and the Oct. 6 coverage of the test said it would not automatically become mainnet's limit ^[3]. The limit is also a client default. Prysm's Oct. 5 release notes show version 7.2.1 adding the Sepolia schedule so validators default to 200 million gas after the fork, while the previous version kept 60 million unless operators changed their configuration ^[9].

Curve's own yardstick is execution. "What we think should be watched most closely is execution under real load," the team said ^[7]. Its practical tests are how quickly transactions enter blocks and what users actually pay during demand spikes ^[6]. It said a higher ceiling offers limited benefit if transactions keep stalling in busy periods ^[14]. It attached one condition: added capacity is welcome when Ethereum can provide it without making validation harder ^[8].

In the case Curve is hoping for, spikes leave headroom, liquidations land on time, and cheaper arbitrage keeps LLAMMA, its liquidation system, closer to market prices as collateral passes through its range ^[11]. The team also expects smaller arbitrage trades and longer swap routes to become worth running ^[13]. In a worse case, demand fills the bigger blocks during a sell-off, as Curve concedes it could ^[10], and the borrower still pays 14.7 swaps' worth of gas at a spiked price ^[17]. The third case touches Curve's own numbers. Glamsterdam reprices state operations, and some actions will need more gas ^[12]. The 1.84 million figure was measured on mainnet before any of that took effect ^[21], and Curve did not say which of its transactions would get heavier.

I think Curve has chosen the right test, or rather the test its own lending book is most exposed to. Liquidations that depend on arbitrage arriving while collateral sits inside a price range gain most from blocks that still clear during a crash ^[11]. The counter-case is that a 3.3-fold ceiling ^[20] gets absorbed on the first volatile day and fees per unit of gas go back to where they were. Curve's comments also stopped short of claiming that every user action would become cheaper ^[12]. If borrow inclusion times and fees on a high-volatility day after activation look like today's, Curve's case for the upgrade fails by its own measure.

Cryptobriefing tied the announcement to October price markets for ether, saying the news was seen as potentially supportive ^[23]. Those markets priced a dip to $2,400 at 82% and a move to $3,300 at 3% ^[22]. The upgrade has been tested on Sepolia and has not been activated on mainnet ^[4].