Curriculum·J313 Mining, Validating, and Running Infrastructure·about 35 min
Validating
By the end of this lesson you can
- →Distinguish the two things a validator is penalized for, and why only one is serious
- →Explain how the correlation penalty makes your loss depend on other people's setups
- →Compute a solo validator's net income against the cost of one slashing event
- →Identify the configuration choices that convert an independent risk into a correlated one
Junior · enrolled learners
This lesson opens with The Staked slashing, 2 February 2021.
- What happened
- Staked, a staking service provider, stopped persisting the Prysm client's on-client slashing protection database across validator client restarts, because the database was a heavy user of disk I/O and was causing missed attestations. Staked relied instead on a separate double-signing protection layer it ran on HashiCorp Consul. It then scaled up its beacon nodes for performance, and under production load its validators restarted more often than they had in testing. Without the persisted database, the restarted validators signed a second version of blocks they had already signed, a conflicting signature that is slashable by design. 75 Staked-run validators were slashed on 2 February 2021, costing about 18 ETH, roughly $30,000 at the time. By Beaconcha.in's data it was the largest single slashing event since the Beacon Chain went live on 1 December 2020. Staked said it would reimburse its clients for both the slashed ETH and the lost rewards.
- The decision point
- The failure was one safety layer removed for performance, on the belief that a second layer was enough, in an operation run by professionals, and the mechanism that caught it was working exactly as specified. Nothing was hacked and nobody acted in bad faith. What made it 75 validators rather than one is that the change was rolled out across the fleet, which per J312-01 turns an independent risk into a correlated one and is the single most consequential decision an operator makes.
- Recorded loss
- $30,000
What you will be able to answer
- →What are the two penalties?
- →What is the correlation penalty?
- →How do the two compare in money?
- →What creates correlation?
Orientation and Year One are open: anyone can read them without an account. From Year Two onward the lessons are for enrolled learners, because progress through the later years only means anything if it is tracked against a record.
It is free. We do not sell the list and there is nothing to buy at the end of it.
Sources and review
- https://web.archive.org/web/20210215083636/https://blog.staked.us/blog/eth2-post-mortem
- https://cointelegraph.com/news/expensive-lesson-75-eth2-validators-slashed-for-introducing-potential-chain-split-bug
- https://www.theblock.co/post/93730/eth2-validators-slashed-staked-bug
- https://beincrypto.com/75-eth-2-0-validators-double-vote-lose-30000-to-penalties/
- https://github.com/ethereum/consensus-specs/blob/master/specs/electra/beacon-chain.md
- https://github.com/ethereum/consensus-specs/blob/master/specs/phase0/beacon-chain.md
- https://eips.ethereum.org/EIPS/eip-7251
- https://blog.ethereum.org/2025/04/23/pectra-mainnet
- https://web.archive.org/web/20260414155145/https://www.blocknative.com/blog/an-ethereum-stakers-guide-to-slashing-other-penalties
- https://medium.com/offchainlabs/post-mortem-report-ethereum-mainnet-finality-05-11-2023-95e271dfd8b2
Confidence medium·Volatility high·Reviewed 2026-08-06·Owner unassigned
Contested
Slashing parameters have changed three times since the incident. In February 2021, under the original (phase 0) specification, the initial penalty was 1/128 of the balance, 0.25 ETH on 32, with a proportional multiplier of 1, which is consistent with the reported total of about 18 ETH across 75 validators. The Altair upgrade in October 2021 moved these to 1/64 and 2. From the Bellatrix upgrade in September 2022 the initial penalty was 1/32, 1 ETH on 32, with a multiplier of 3. Since the Pectra upgrade on 7 May 2025 it has been 1/4096, under 0.008 ETH on 32, with the multiplier still 3; EIP-7251 cut it to make consolidating stake into larger validators less risky. Part one states the Pectra mechanism, and the autopsy reports the 2021 total as reported. Both the quotient and the multiplier are protocol parameters that can change again, so re-derive rather than assume.
The 36 days of missed-duty penalties in part three are 8,192 epochs, the period after which a slashed validator becomes withdrawable. Valuing those penalties at roughly the reward forgone is an approximation: under the current rules a validator that misses an attestation loses the source and target weights, 40 of the 54 weight units a timely attestation earns, about three quarters. The same approximation is used for ordinary downtime in part four, so the comparison between the two is like for like.
The validator economics in part three use stated assumptions, being a 3.0 percent annual reward rate, an ETH price of $3,000, a $1,200 machine amortized over five years and power at $0.15 per kilowatt hour. Reward rates move with total stake and none of these figures should be treated as current.
The Staked event is dated 2 February 2021 by Staked's own post-mortem, published the following day. Cointelegraph's report of 5 February dates the slashings to 4 February and lists them as attestation violations, while Staked describes the restarted validators as signing a second version of blocks they had already signed. The lesson follows the operator's account for the date and the mechanism and does not name the message type.
This lesson covers running a validator yourself. J313-03 covers handing that job to somebody else. Keep the split and route custody of the withdrawal key to F103.