End a snooze now, so paging continues from the current step

Why a page turned out to happen.

A fixed list rather than free text, so the next firing of the same rule can say "3× config change". Free text fragments into near-duplicates and never groups. Nuance goes in cause_note.

The sentence beside the structured cause. One dropdown plus one line is the whole ask — anything longer does not get filled in at 3am.

When the ladder ran out with nobody answering. None while it is still climbing, or if somebody took it first.

Deliberately NOT a ResponseState. An exhausted page can still be acknowledged and resolved, so a lifecycle state would force a false either/or: exhaustion is a property of the ladder, the state is a property of the human.

Set only when this firing also produced an incident. A record is never owned by an incident, so this stays None for most firings.

Where step delays count from: opened_at, moved by a snooze or a triage-held first page.

Which run of the ladder the record is on.

Handing a page to another person or team starts a new run, and the ledger is read per run — which is what makes the receiving responder's ladder begin at its first rung. Absent means the first run.

For an impacted record, the owner record it was opened alongside.

AlertPriority::to_i32 — the same scale alerts already use.

Why this team was paged.

A database goes down and five services break. The owner fixes the database; the impacted teams contain the blast radius on their own service. Different jobs, so they get different records, and each team acks and resolves its own.

Quiet until this instant, in micros.

Not an acknowledgement: snoozing says "I know, stop shouting", not "I have this". The record stays open and unowned and the ladder resumes when it expires — which is what makes it safe to offer.

Lifecycle of a response record: triggered → triaged → acknowledged → resolved. triaged comes from the L0 agent and is skipped entirely when the agent is absent or disabled.

The team that owns this firing, or None when routing found nobody.

A teamless record pages nobody — no policy, no rotation, no ladder — and exists so the firing appears beside the ones that did page. Handing it to a team gives it an owner and starts its ladder (R-D6).

What the page is about. Kept on the record so it survives the alert being renamed or deleted.

Revision of the write this record came from, and what a replica accepts or refuses a snapshot on. The queue redelivers an unacked message behind newer ones, so a blind apply lets an older snapshot un-acknowledge a page that a human already answered.