APP status remains "active" even after agent is lost

[Nobuto Murata]

How to reproduce:
 1. deploy one service
 2. forcibly stop the machine hosting the service (lxc stop --force or IPMI force power-off)

Expected:
 APP status will be changed from "active" to something indicating error

Actual:
 APP status remains as "active"

$ juju status mysql
MODEL CONTROLLER CLOUD/REGION VERSION
default localhost lxd/localhost 2.0-beta16

APP VERSION STATUS EXPOSED ORIGIN CHARM REV OS
mysql active false jujucharms percona-cluster 2 ubuntu

RELATION PROVIDES CONSUMES TYPE
shared-db glance mysql regular
shared-db keystone mysql regular
cluster mysql mysql peer
shared-db mysql neutron-api regular
shared-db mysql nova-cloud-controller regular

UNIT WORKLOAD AGENT MACHINE PUBLIC-ADDRESS PORTS MESSAGE
mysql/0 unknown lost 2 10.0.8.122 agent is lost, sorry! See 'juju status-history mysql/0'

MACHINE STATE DNS INS-ID SERIES AZ
2 started 10.0.8.122 juju-4a9c27-2 xenial

[Anastasia (anastasia-macmood)]

This issue has been addressed in recent times as part of other observability related issues.

I have recently tested it with AWS:
* bootstrap;
* deploy charm with some units;
* stop a unit machine via CLI (outside of Juju);
* watch juju status.

Machine status was updated to 'down'.