Observed on hanzo-mainnet (36963), v1.36.56/57/58, 2026-08-05/06.
A node with a wiped data dir processes cert/context catch-up frames endlessly with processed context entries ... certAccepted=0 certRejected=0 voted=256 — thousands of entries processed, zero adopted — and requested context for missing prerequisites loops forever. Three resync attempts stalled at heights 892, 1241, 3915 (of 6698) while OOM-killing the serving peers (companion issue).
Behind-group nodes in a frontier split show the same signature: they can never adopt the ahead-group's certified blocks live (re-poll: gossiped block abandoned after attempt cap — not re-soliciting further (recoverable via cert/catch-up) — but the recovery path adopts nothing).
Working recovery was a state TRANSPLANT (CSI snapshot of a healthy peer's C-chain dir onto the wiped node's PVC — excluding vote-guard, which is per-node) — sync-by-protocol was unusable.
Observed on hanzo-mainnet (36963), v1.36.56/57/58, 2026-08-05/06.
A node with a wiped data dir processes cert/context catch-up frames endlessly with
processed context entries ... certAccepted=0 certRejected=0 voted=256— thousands of entries processed, zero adopted — andrequested context for missing prerequisitesloops forever. Three resync attempts stalled at heights 892, 1241, 3915 (of 6698) while OOM-killing the serving peers (companion issue).Behind-group nodes in a frontier split show the same signature: they can never adopt the ahead-group's certified blocks live (
re-poll: gossiped block abandoned after attempt cap — not re-soliciting further (recoverable via cert/catch-up)— but the recovery path adopts nothing).Working recovery was a state TRANSPLANT (CSI snapshot of a healthy peer's C-chain dir onto the wiped node's PVC — excluding vote-guard, which is per-node) — sync-by-protocol was unusable.