Release v19.28.0

Added

• Chronicle records an alert in its system event sequence when a partition keeps failing past the grace period, escalates it when the partition runs out of retries, and raises one when an observer is quarantined. Each incident is cleared when the partition recovers or is cleared, or when the quarantine ends. The Alerts configuration section now takes effect. Delivering alerts to email or chat comes later. Understand recorded incidents describes what is recorded and when. (part of #4424)

Changed

• Retiring a projection and removing an observer now clear the observer's open incidents first, and fail if they cannot. Retrying the operation continues from where it stopped: a retired observer stays retired, and automatic recovery after a restart no longer reactivates it. Registering it again starts it fresh. (part of #4424)