Release Notes - v0.72.0

v0.72.0

Title: Strengthen validation, synchronization, and operational efficiency

Date: 2026-09-24

Version: v0.72.0

Release: v0.72.0

This release strengthens ledger, governance, network, and credential validation while improving synchronization, reward processing, database durability, and operational visibility.

Recommended Network Compatibility ⚠️

Network	Compatible
mainnet	⛔
preprod-testnet	✅
preview-testnet	✅
musashi	✅

Validation and compatibility improvements

• Enhanced peer sharing and made pool relay decoding stricter through gouroboros v0.206.2; oversized SharePeers replies now fail strictly, and governance, ledger, and threshold cache behavior receive the upstream correctness improvements.

• Rejected Byron outputs whose network magic differs from the active network, preventing outputs from another network from entering validation.

• Corroborated chain frontiers that advance beyond the catch up cap with distinct connections before accepting them, reducing the risk of following an unverified frontier.

• Reduced SQLite durability overhead for the Koios parity cache and test databases by using synchronous=NORMAL for the cache and synchronous=OFF for test-only connections; disposable test caches can opt into relaxed durability.

• Blocked duplicate Byron and Shelley genesis UTxO references before the ledger writes partial state, keeping invalid initial state out of the ledger.

• Preserved replaced governance votes across rollback through vote history migration, maintaining the information required to reconstruct governance state.

• Restored fail closed NewConfig defaults for historical validation and strict UTxO validation, so those checks remain enabled unless explicitly changed.

• Clarified strict Koios parity handling so epochs affected only by reference_lag do not shut down parity checks, while genuine data failures remain fatal.

• Aligned Byron canonical JSON handling with the reference behavior for duplicate keys, escape sequences, and unsigned fields.

• Evaluated SPO governed actions against the stake snapshot at the same boundary and prevented actions from appearing ratified when they did not pass ratification.

• Protected Mithril credentials and signed URL data through redaction, and rejected URLs that contain user information.

• Validated LocalStateQuery point queryability during Acquire, improving pinned era checks and retention handling before Dingo uses a point.

Synchronization, governance, and observability

• Repositioned gosec G703 lint suppression comments so security checks continue to pass without changing runtime behavior.

• Removed the retired Go Report Card badge from the project presentation.

• Switched the Antithesis Moog download to GITHUB_TOKEN authentication.

• Expanded operational metrics with epoch_rollover timing, stage maximum gauges, wider latency buckets, and dashboard panels.

• Introduced Koios backed node parity validation from genesis with complete and incomplete result reporting.

• Hardened Byron witness validation by rejecting malformed or unknown witnesses.

• Relaxed MySQL and PostgreSQL durability settings for CI test services.

• Batched reward live stake rebuilds with bounded memory use and ETA progress reporting.

• Corrected reward accounting so the active stake denominator uses the full delegation set identified from credentials, including stake delegated to inactive or retired pools, while reward buckets and reward_pool_input rows remain limited to active pools. Reward snapshots now preserve this stake in reward_snapshot.total_active_stake and record exclusions in reward_snapshot.excluded_active_stake for degraded registrations and inactive or retired pool delegations. Reward round logs and precompute attribution, including reused precomputations, now report reserves, fees, total_active_stake, total_circulation, total_blocks, and reward_efficiency.

• Updated plutigo to v0.7.2, preventing a Preview reference script validation halt while retaining execution safety.

• Enforced Shelley family DELEG and MIR certificate predicates during validation.

• Classified missing redeemers as deterministic failures so replay recovery can handle them consistently.

• Stabilized a blockfetch refusal test by removing a race in its promotion handling.

• Pipelined blockfetch range requests during deep catch up and improved request and error handling.

---

Docs authored by Doc Holiday