[
  {
    "mechanism": "physical streaming standby",
    "useful_when": "A current PostgreSQL endpoint must remain available after an instance failure or serve compatible read traffic.",
    "owned_obligations": "Operate authentication, WAL retention, replay monitoring, promotion, timelines, and application acceptance after failover.",
    "reconsider_when": "The failure to cover is logical deletion or corruption which has already reached the standby."
  },
  {
    "mechanism": "independent snapshot",
    "useful_when": "A storage or provider boundary can retain a consistent historical point with suitable restore granularity and duration.",
    "owned_obligations": "Verify consistency, identity, retention, permissions, restored configuration, and application behavior.",
    "reconsider_when": "The available points are too coarse, too recent, incompatible, or too slow for the declared recovery contract."
  },
  {
    "mechanism": "physical base backup and WAL archive",
    "useful_when": "A compatible PostgreSQL cluster needs a selectable historical point between base backups.",
    "owned_obligations": "Retain a valid base, complete WAL chain and timelines, choose the right target, and test application acceptance.",
    "reconsider_when": "Compatibility, archive completeness, operator capability, storage cost, or recovery duration cannot meet the declared contract."
  }
]