Release v19.32.0

Summary

Chronicle now keeps a queryable record of open observer alert incidents. A kernel reactor materializes the recorded raise, escalation and clear transitions into one row per incident, and three scoped queries let operators list, look up and count open incidents for an event store, together with whether the incident view is up to date.

Added

• Open alert incidents are materialized from the recorded alert transitions into incident rows stored next to the System event sequence, for MongoDB, SQL (SQLite, PostgreSQL, SQL Server) and in-memory storage. Redelivered or out-of-order transitions never overwrite newer state, and a cleared incident cannot be resurrected by a delayed raise (part of #4424)
• GetOpenIncidents, GetOpenIncident and GetOpenIncidentCounts queries on the new Alerts service, scoped to an affected event store and optionally a namespace, with filtering by observer, condition and minimum severity, bounded keyset paging and counts grouped by namespace, condition and severity (part of #4424)
• Every incident query reports whether the incident view is Ready, CatchingUp or Degraded, so an empty result is never mistaken for "no incidents" while the view is still catching up or has failed partitions (part of #4424)
• Documentation for querying open alert incidents: https://www.cratis.io/chronicle/hosting/alert-incident-queries/ (part of #4424)

Changed

• Generated shared enums in the contracts now carry the documentation of their source enums, and optional enum query arguments are supported in generated requests (part of #4424)