It's not a matter of good or bad developers. A full upgrade requires to stop what's running. It takes one day probably because there's a lpt tp do in the meanwhile, like taking snapshots and backups of the system, upgrading and then testing the upgraded machine back again.
As long as you have servers in place, you'll need shutdowns. "Serverless" is bullshit, it's simply someone else's server.
I agree that bitcoin solves that specific thing, but custodial services on top of bitcoin that still use servers and databased still have to do shutdown maintenance. Nostr Relays run on servers...upgrading the posgres db will need to stop it, otherwise you could face some corruption issues.
C'mon