Skip to content
Data provided by CoinGecko
Ethereum News

Reth v2.11.0: high-priority RPC caps, not a fork

Comic illustration of a mittened beetle watching a glowing crystal node

Paradigm tagged Reth v2.11.0 at 2026-11-11T10:47:00Z as a hardening and performance release. The notes mark update priority High for payload builders and High for non-payload builders. This is an execution-client binary. It is not an Ethereum mainnet fork.

Ether (ETH) traded at $2,056.35 on the CoinGecko snapshot fetched at 2026-11-12T00:07:43Z, up 0.27% over 24 hours. Bitcoin (BTC) was $73,971 (up 0.36%). The global crypto market cap sat near $2.479 trillion, with Bitcoin dominance about 59.92%. Price does not tell you whether a node still accepts an unbounded eth_callMany. The tag does.

I care about shipped client mechanics more than roadmap slides. Show me the fee, and show me the binary. v2.11.0 is the binary.

What happened

The GitHub release is not marked as a pre-release or a draft. The summary says RPC calls that can do unbounded work are now capped or cancellable: eth_callMany, trace_callMany, debug_traceCallMany, eth_getProof, mev_simBundle, and eth_simulateV1. It also lists transaction-pool, peer-to-peer, and storage correctness fixes, plus parallel EIP-7702 authorization recovery and cheaper overlay provider setup.

The action-required block is operational, not rhetorical. eth_callMany, trace_callMany, and debug_traceCallMany now allow at most 1,024 calls and share one --rpc.gascap budget across the request. No additional call starts more than five seconds after the first. Previously successful requests may now fail. eth_getProof allows at most 1,024 storage keys. eth_getMultiProof allows at most 1,024 targets and 1,024 storage keys in total. Requests over those limits fail with invalid params.

Operators who still pass retired no-op flags will not get a warning toast. The node refuses to start if the config still includes --engine.legacy-state-root, --engine.caching-and-prewarming, --engine.parallel-sparse-trie, --engine.disable-parallel-sparse-trie, --engine.precompile-cache, or --rpc-cache.max-headers (alias --rpc-cache.max-envs). With --instance N and no explicit --auth-ipc.path, the Engine API IPC socket is now /tmp/reth_engine_api.ipc-N. Set the path if a consensus client still expects the old socket. Explicit IPC socket permissions are validated at startup. An invalid mode is a startup error, including on platforms that cannot set the mode before binding.

The same notes still carry a backwards-compatibility caveat: database migration, resync requirements, and supported downgrade versions “must be confirmed against the final v2.11.0 release build before publication.” That sentence is still on the published tag. Treat it as unresolved in public, not as a silent all-clear. Pre-built Linux and macOS binaries are listed and signed with PGP fingerprint 50FB 7CC5 5B2E 8AFA 59FE 03B7 AA5E D56A 7FBF 253E. Verify the signature. Do not skip that step because the RPC caps look tidy.

Context

Reth’s own update-priority page (docs currently labeled v2.11.0) uses the same High / Medium / Low table language as Lighthouse. High means the update should be completed as soon as possible, on a hours-or-days clock, for payload builders and for everyone else who runs the binary for RPC or apps. Medium is days or a week. Low is the next routine cycle. v2.10.1, tagged 2026-11-06T18:53:31Z, listed Medium for both classes. v2.11.0 moved both rows to High. That is a priority change in the notes, not a marketing adjective.

The fee tape is quiet next to that table. DefiLlama’s Ethereum fees overview, read at 2026-11-12T00:10:12Z from api.llama.fi/overview/fees/ethereum, showed about $10.07 million in fees over the last 24 hours, down 2.35% versus the prior day, with about $78.92 million over seven days. Ethereum TVL on DefiLlama’s chain list at the same read was about $41.38 billion. Quiet gas fees do not lower the cost of leaving a public RPC method unbounded. They also do not turn a client tag into a protocol fork. For how those prints sit next to price, the desk’s market snapshot guide is the method. For the asset’s cycle, see Ethereum history.

SDK users get a separate breaking path. Ethereum block execution now runs on evm2::block, with Reth executor types as thin adapters. Discovery dependencies moved to discv5 0.12, ENR 0.14, and secp256k1 0.33. Those lines matter if you compile against Reth as a library. They do not rewrite mainnet rules. The P2P section is closer to operators who expose a node: banned and untrusted inbound peers are rejected before session activation, unsolicited responses are no longer decoded, malformed disconnect payloads are bounded, and several pool and storage races are closed. That is hardening of a live execution client, in the same family as the recent Lighthouse v8.3.3 high-priority patch, not a scheduled hard fork on the Ethereum news calendar.

Our read

Ship the binary if you run Reth for payloads or for public RPC. Do not wait for a fee spike to make the High label feel urgent. The notes already spent the urgency on both user classes. The 1,024-call cap and the shared gas budget are the parts that will show up first in logs: requests that used to complete can now fail with invalid params. That is the point of the release. An unbounded simulation batch is not a feature. It is a way to pin a node.

Do not confuse this tag with Glamsterdam, Hoodi, or any other network clock. v2.11.0 does not change consensus. It changes how this client spends CPU and memory on RPC, proofs, and gossip. Keep the migration caveat in view. The published notes have not yet replaced “verification pending” with a downgrade matrix. Confirm that line against the tag you actually download before you treat the upgrade as a no-resync drop-in.

Falsifiable claim: By 23:59 UTC on 2026-11-19, the published v2.11.0 GitHub release notes will still list Update Priority High for both Payload Builders and Non-Payload Builders. If Paradigm edits either class to Medium or Low, or states that the High labels were a copy error, this ship-now read is wrong.

What to watch next

First, whether operators actually drop the retired CLI flags. A node that still passes --engine.legacy-state-root will fail to start. That is a config break, not a chain halt. Watch startup logs, not Twitter forks.

Second, the Engine API socket path on multi-instance hosts. If a consensus client still points at the unsuffixed IPC path, the pair will not talk after --instance N. That shows up as missed payloads, not as a new ticker.

Third, the backwards-compatibility sentence. If a later edit of the same tag, or a follow-up patch, states a clean downgrade path to v2.10.1 or names a required resync, update the runbook. Until then, do not invent one.

Fourth, RPC client breakage. Indexers and MEV simulators that send eth_callMany batches above 1,024 calls, or proofs with more than 1,024 storage keys, should fail closed. If they keep succeeding on a node that claims v2.11.0, you are not on the binary you think you are.