Arbitrum vs Base

Latency is conditional on success: the harness times only calls returning a valid fresh block, so a mostly-failing endpoint can post a fast p50. Rows below 50 percent success are degraded; unresponsive probes pin below the table. The All chains aggregate favours single-chain foundation endpoints (Base, Binance, Arbitrum, Optimism, Avalanche officials) since they rank on one chain only; use a chain tab for honest comparison. Dead endpoints (cloudflare-eth POST returns -32603) are excluded.

A zero row means no incident was caught in the window, not a proof of correctness. Most providers post zero most days; this bench exists for the days they do not, and for the errors: a blocked eth_getLogs or a balance read gated behind a paid key at 40 blocks of depth counts as an incident because it is one from the caller's seat.

Bridged TVL is value secured by the chain's bridge, not value locked in its DeFi protocols; a chain can rank high on one and low on the other. The split is L2Beat's: canonical is escrowed on the host chain by the chain's own contracts, external came through a third-party bridge, native was minted here. Rows are the chains L2Beat tracks in its scaling summary; a settled L1 has no row by construction, not by omission. Size is not quality: a large balance says capital came, not that it stayed.

Stablecoin float is where dollars are parked, not what they do. A chain can take in a billion that sits idle, and lose float to a single issuer redeeming. The figure is DefiLlama's chain-wide pegged-USD aggregate, every issuer combined; it does not separate a bridge inflow from a native mint, and a dollar bridged to an L2 counts on both chains, so rows are not additive. Read the net dollars with the percentage beside them: the same $400M is a rounding error on Ethereum and a doubling elsewhere.