Overview
SDKs evaluate flags locally and keep their last good snapshot when refresh or streaming fails. Flag changes publish automatically, with retries for failed publication. See Local evaluation for startup defaults and outage behavior.Backups and restore
The standard database backup target is one encrypted off-site dump every 24 hours with 35-day retention, plus a daily automated restore verification. Shipd manages database recovery and republication of flag configuration. These are engineering targets, not a contractual SLA; ask support for the latest drill evidence and your contracted RPO/RTO. Extended retention and point-in-time recovery are not generally available.Incidents and rollback
During an incident, check the service status/support channel, stop repeated writes, preserve request and approval ids, and keep SDKs on their last good snapshot. Flag rollback is a new audited change: revert the config, runshipd diff, then shipd push; an approval policy may queue it. Platform rollback and restore are operator actions. Never bypass an approval or force a conflict unless your incident procedure authorizes it.