Recovery expectations
What may recover quickly and what must wait for reconciliation.
Recovery speed depends on which layer failed. SV3 does not restore writes merely because a process restarted.
| Failure | Expected behavior | Required recovery evidence |
|---|---|---|
| Static web | Direct contract and explorer access can remain available | Pinned asset digest, domain and release verification |
| API or RPC | Reads and submission may be unavailable or fail closed | Correct chain, current head, code hashes, readiness and error recovery |
| Indexer or database | Charts, history and portfolio views become stale | One-writer backfill, readiness, reorg checks and zero unexplained reconciliation delta |
| Keeper | A new floor raise may be delayed; existing floor state remains | Fresh on-chain plan, nonce/queue reconciliation and bounded gas funding |
| Contract or accounting alert | State-changing actions may be fully paused | Independent root cause, reserve/supply/debt/collateral reconciliation and reviewed remediation |
| Reserve asset or chain | Transfers or all execution may be unavailable | Independent provider/issuer facts, current chain state and an accountable resume decision |
For local/tabletop planning, critical alerts target acknowledgement within five minutes, incident command within ten minutes, and a pause or degraded-mode decision within fifteen minutes. Those targets do not justify unsafe retries, manual database edits, an unreviewed upgrade, or an unexplained accounting delta.
Source or artifact remediation can invalidate previous release evidence. In that case the changed code and artifacts must be rebuilt, tested, reviewed, and authorized again before promotion.