Ethereum was not created to streamline finance or make apps convenient. It was designed to set people free.
This line in the Trustless Manifesto drew criticism at the time of its publication, and Vitalik Buterin reiterated it on January 5th.
Discussion: Ethereum’s mission is fundamentally different from the efficiency game in which DeFi protocols compete. The goal is not to reduce yields from 4.5% vs. 5.3%, or reduce latency from 473ms to 368ms, or reduce signups from 3 clicks to 1 click.
Ethereum’s game is resilience. It’s about avoiding overall loss if the infrastructure collapses, the government becomes hostile, or the developer disappears. Resilience means that 2,000ms of latency remains 2,000ms even if Cloudflare fails, a sponsor declares bankruptcy, or a user is removed from the platform.
Resilience continues to be a first-class participant, regardless of geography or politics.
This is important because Ethereum locks in nearly $74 billion in smart contract value and over 65% of tokenized real-world assets in Layer 1 alone.
But the system designed to be the world’s computer rests on a surprisingly fragile stack of centralized chokepoints.
The consensus protocol continued to finalize blocks, but an outdated client of the RPC provider caused the exchange to crash. The blockchain continued to operate, but the CDN went down and half of the ecosystem went offline.
Preventing disaster rather than optimizing yield
A recent report quantifies that risk. Infrastructure failures cause 5.7 times larger volatility shocks than regulatory announcements across major crypto assets. Tail risks such as complete loss of access, permanent lock-up of funds, and network outages outweigh increased revenue.
A protocol that provides a 5.3% yield is worthless if a configuration error can destroy your infrastructure.
Vitalik Buterin’s composition captures this. Resilience isn’t about how fast everything works, it’s about whether your application will run at all if your infrastructure provider disappears or your hosting platform leaves your platform.
Ethereum’s 2,000ms latency may be slower than Web2, but it will continue to deliver even if the Web2 system goes completely down.
Still, Ethereum’s promise of resilience faces practical tests.
In November 2020, Infura, the default RPC provider for MetaMask and most DeFi apps, ran an older Geth client that was forked from the canonical chain.
Exchanges stopped Ethereum withdrawals, Explorer showed inconsistent states, and MakerDAO and Uniswap broke for users.
Although the bugs themselves have been fixed and alternative RPC implementations are in progress, centralization is still the norm. It’s not just Infura, it’s more of a “small cartel”.
The protocol worked, but the connection point failed.
In November 2025, a Cloudflare configuration error brought down approximately 20% of our web traffic, including Arbiscan, DefiLlama, and multiple Exchange and DeFi frontends. Ethereum continued processing blocks. User could not access.
During the epitaph boom of 2024, Arbitrum’s single sequencer was down for 78 minutes. No transactions were processed and no batches were posted to Ethereum.
Arbitrum, Optimism, Base, and zkSync all currently rely on a single centralized sequencer. Although the decentralized base layer worked correctly, the centralized infrastructure prevented users from benefiting.
| layer | current dependencies | Vulnerability indicators | resilient alternatives |
|---|---|---|---|
| Access/RPC | Infrastructure, Alchemy, Quicknode; MetaMask default is Infura | Up to 90% of Web3 app traffic. November 2020 Infura outage halts ETH withdrawals and breaks MetaMask, MakerDAO, and Uniswap | Comes standard with multiple RPC providers, local light clients, and stateless clients. RPC diversity as a user-facing feature |
| relay/builder | MEV-Boost relays (Ultra Sound, Titan, bloXroute) mediate over 90% of blocks | Four relays control >85% of proposals. Titan, Beaverbuild, Rsync generate >80% of builder blocks | Increased relaying by separate entities. Relay neutrality. Holy PBS, Blockspace doesn’t stop due to relay failure |
| L2 sequence | Single Sequencer (Arbitrum Foundation, Optimism Foundation, Coinbase for Base) | Arbitrum: 78 minutes of downtime. Base gets 70.9% of L2 profit, Arbitrum gets 14.9%, Optimism gets 5.4% | Distributed sequencer set or L1 fallback. Force inclusion if the sequencer censors it. Tracks under single control % L2 TVL |
| DNS/CDN | Cloudflare for DNS, TLS, and dApp caching | Cloudflare powers about 20% of the world’s web. Arbiscan, DefiLlama, and exchange/DeFi frontends will be down due to service outage in November 2025 | IPFS/Arweave with ENS fallback. Multi-CDN. Wallet invocation contract without web front end |
| Basic protocol | Ethereum consensus (Lighthouse 52.65%, Prysm 17.66%). Execution (Geth ~41%, Nethermind 38%) | September 2025 Reth bug brings down 5.4% of nodes. Diversity precluded widespread impact. | No client has a share greater than 33%. home staking. Minimize related failures. Easy light/stateless client validation |
The underlying protocol exhibits true resiliency with multiple clients, hundreds of thousands of validators, and proof-of-stake that spreads risk across different codebases.
When Reth encountered a bug in September 2025, 5.4% of its nodes went down, but Geth, Nethermind, and Besu continued, so network continuity was maintained. Client diversity worked.
The problem is centered on the above. RPC access, relays, sequencers, and web front ends introduce dependencies that disable user access even when the base layer is functional.
This is where Ethereum’s resilience breaks down. It’s not about encryption or consensus, it’s about the scaffolding that connects users to the protocol.
Centralized sequencer as an economic bottleneck
Layer 2 sequencers centralize both control and benefits. Base consistently captured more than 50% of all rollup profits throughout 2025, followed by Arbitrum.
The Arbitrum sequencer is operated by the Arbitrum Foundation, the Optimism sequencer is operated by the Optimism Foundation, the Base sequencer is operated by Coinbase, and the zkSync sequencer is centrally managed.
As a result, more than 80% of the fees collected by Ethereum’s Layer 2 in 2025 flowed to blockchains with centralized sequencers.
A technical path exists. A shared sequencer network like Espresso, or a rollup based on returning sequences to an Ethereum validator. Astoria attempted a similar design, but closed in 2025.
The gap is economic, not technological. A centralized sequencer provides a better UX and generates significant revenue. To increase resiliency, you must accept that a sequencer that cannot be shut down by a single operator will outweigh the millisecond improvement of single-point control, although it will be slightly slower to confirm.
RPC and CDN dependencies
MetaMask defaults to Infura. According to the report, most Web3 applications use Infura, Alchemy, or QuickNode.
The Infura incident in November 2020 demonstrated that protocol-level resiliency becomes meaningless when the access layer fails.
Cloudflare’s November 2025 outage revealed how dependent “decentralized finance” is on one company’s CDN. Ethereum processed the block successfully, but users were unable to access the frontend, explorer, or dashboard.
Resilient alternatives include wallets that use multiple RPCs by default, local light clients, distributed storage on IPFS or Arweave, ENS addressing, and multi-CDN deployments.
However, these come at a cost, such as increased complexity, increased bandwidth requirements, and increased management complexity.
Most projects choose convenience, so trade-offs with efficiency are important. Ethereum’s base layer provides survival properties, but the ecosystem almost always wraps them with dependencies that reintroduce any vulnerabilities.
the actual trade-off
The value proposition for Ethereum that Buterin frames is not faster, cheaper, or more convenient. It works even if everything else is broken.
This requires infrastructure choices that prioritize survival over optimization. This means multiple client implementations if technically better, diverse RPC providers if latency is better, distributed sequencers if a centralized operator provides faster confirmations, and distributed frontends if centralized hosting is easier.
The industry has not accepted this trade-off. Rollups optimize the UX and accept the risks of a single sequencer. Applications use convenient RPC by default and accept concentration risks. The frontend is deployed on a commercial CDN and tolerates single vendor failure.
The choice is to build in case Cloudflare, Infura, and Coinbase all continue to work, or build in case they don’t.
Ethereum’s base layer allows for the second option. The surrounding ecosystem overwhelmingly creates the first.
The protocol provides 2,000ms of latency and persists through infrastructure failures, platform decommissions, and geopolitical turmoil.
Whether someone builds a system that actually leverages that property, rather than wrapping it in dependencies that reintroduce all the vulnerabilities that Ethereum was designed to eliminate, will determine whether resilience becomes a reality or remains theoretical.
Block space is plentiful. A decentralized, permissionless, resilient blockspace is not.