Hi oscarf,
In general, the best practice w/ Juju is to monitor it using the same LMA (Logging Monitoring Alerting) stack that you use to monitor deployed models. There are hooks in Juju that make this possible. Here’s a writeup on the subject with some specific examples.
Some of the guidance on monitoring charmed kubernetes might also be useful.
(Note to other members of the Juju team: the above should probably be incorporated into the core documentation.)