This allow the person responsable on for support, load all monitoring URLs from all instances to that may contain a failure.

Stop or Destroyed instances don't required checks.

This cause a lot of headeache during the tests.

/reviewed-on !48

on generated supervisord config

If a promise anomaly timeout or return no message, we return error if
the problem is happening for a second time after a second run or if there was already an error.
This will reduce temporary failure.

On a setup for private cloud (internal master), reguralisation requests are not relevant and risky as alarms may trigger removal of instances due unpaid invoices.