Mandatory upgrade before H12
H12 moves to DAA 92 550 000 (≈ Saturday 2026-09-05 16:00 UTC). Upgrade before the pruning sample of Thursday 2026-09-03 04:00 UTC to be certain, and no later than Thursday ~21:00 UTC.
Fixes
- Fresh sync after H12 was broken. A node syncing from an empty datadir could not use the production snapshot it imported (foreign chain numbering, no history below the pruning point) and panicked at the first validated block. The snapshot now carries the daa of every chain index in its window (~200 KB on mainnet); a fresh node verifies exactly from the first block and rebuilds the next samples identically.
- Batch validation could miss a reward finalized inside the batch (IBD or catch-up after downtime), disqualifying a valid block. Hidden so far by the catch-up trust window. Virtual resolution now stops at the ledger's finality frontier. No throughput impact.
- Stale pruning point during validation fell back to chain index 0 on pruned nodes, same panic as the 1.5.8 crash. Diagnosed by @fearke85 in #38, co-author of the fix.
Changes
- Snapshot encoding v2: its hash and the H12 commitment change. Persisted snapshots still in retention are rebuilt at startup, the others dropped. Hence the new gate: its pruning point must be a sample built by 1.6.0 nodes.
- P2P protocol 14. Handshake floor 1.6.0 at the gate (24 h ban below).
Upgrade
In place, no datadir reset. Samples fall at ~04:00 and ~16:00 UTC; the Thursday 04:00 UTC sample is the gate's pruning point. Upgrade before it, or by Thursday ~21:00 UTC at the latest (startup rebuild). Later upgraders follow the chain but get no templates for up to two days. 1.5.9 and 1.6.0 peer together until the gate.