We have changed the downtime model in Horizon 25.1.2 and Meridian 2019.1.2 and haven't documented it accordingly. This issue is related to . The section about "Downtime Model" in the documentation should be fixed. I've found the behavior described in the issue as the following:

Attribute that determines if service is to be deleted when down continuously since the start time. Use 'never' to never delete services, 'managed' to only delete the service if it is actively managed, or 'always' to delete even if the service is on an unmanaged interface.