From public security advisories, bug bounty disclosures, postmortems, audit reports, and maintainer discussions, build a deduplicated table of distinct security incidents affecting the Ethereum consensus clients Prysm, Lighthouse, Teku, Nimbus, Lodestar and Grandine, 2022 through 2026 — one row per incident, with client, versions, date, source type and status, attack class, attacker prerequisites, faithfully stated impact, fix reference, primary-source URL and a supporting quote. Public information only; leave unsupported fields null; do not infer undisclosed vulnerabilities.
The collection holds 47 conservatively screened public records dated March 2022 to September 2026, drawn from 25 independent hosts (the largest host supplies 7 rows, 14.9%). After editorial deduplication of plainly identical events, 43 distinct records remain below. They span confirmed incidents (16), bounty disclosures (9), maintainer advisories (6), audit recommendations (3) and open discussions (9) — most rows are not exploited attacks. Grandine has only two public records here; that is sparse coverage, not evidence of absence. Scan conducted through 2026-10-12.
| Date | Client | Incident | Status | Type | Attack class | Affected / fixed | Prerequisites · impact · fix | Source |
|---|
Method: 47 public records (security advisories, bounty disclosures, postmortems, audit reports, maintainer and specification discussions) covering Prysm, Lighthouse, Teku, Nimbus, Lodestar and Grandine, dated 2022-03-02 to 2026-09-03, gathered from 25 independent hosts; scan conducted through 2026-10-12, public information only. Four records were consolidated as duplicates of the same events: the second May 2023 Lighthouse BlocksByRange crash report (a secondary press cross-post), the Prysm v7.0.1 release note and a later third-party retrospective of the December 2025 Fusaka mainnet incident, and the Nimbus v26.2.0 fix note for the February 2026 mainnet fork (its version and fix details were retained on the postmortem row), leaving 43 distinct rows. Source wording for impact and prerequisites is preserved verbatim; null means the source did not state it. Counts measure public records, not exploitation; no current-code assessment and no inference about undisclosed vulnerabilities. Long quotes are truncated in the table for space — follow each source link for the full text.