Thanks a lot for the detailed answer! I do want to make it clear that I think the approach of a blackholed monitoring canister which is a controller (building on Paul’s blackhole) is a great solution for almost everything and I of course fully trust it works, I just think it’s not a great fit for NNS-controlled things because it makes it harder for people to verify that something is NNS-controlled.
My bad, I forgot about that discussion on upgradability. The named-callbacks are still a planned feature, but we have not made a ton of progress on it yet. The only alternative would be to allow a canister to expose it’s canister status (like was discussed in NNS Proposal: Make canister_status public to anyone).