Yes, It was scary.
I will now read your recommendations at Managing the LXD snap and see which strategy would be best for my client.
I have two other clusters which I have already channelled to 4.19/stable but this problem could not have had been avoided with even that.
It is the first time in running various LXD clusters over few years that I have run into this kind of issue.
I will also have to evaluate to break the cluster and manage each of the nodes separately. Will lose a bit of functionality but not complicated to duplicate in another application level layer.
kind regards