The one down side of the system is the cost. It's cheap to host a PDS but expensive for other components. Users could not relies on "someone" for running those components for free forever.
The difference is that it's easy to scale ATProto AppViews up beyond the reach of their users. It can scale down though. It is not easy to scale ActivityPub instances up beyond the scope of the people who use it, and it would probably be way more expensive if you tried.
- Relay as an optimization. That's cheap-ish ($30/mo) or free-ish if you pool with others.
- Your own app server. That's on par with normal web apps as long as you can keep a socket open. What's expensive is if the app you're making is a fully capable copy of Bluesky itself with gigabytes of existing posts — but is that the app you're making? The economics here are identical to normal web app stuff.
This is the implicit idea in my comment when I said it's expensive. If Bluesky banned me and I could not find another AppView with comparable reach and audience, I lose, the ban is effective.
Another problem is ATProto users don't usually associate then with an AppView the same way Mastodon users associate with the instance, it's hard to raise fund to sustain the infrastructure cost.
Okay, but that's always the case when you get banned from a service. The difference here is that it's possible to build a competing service with different moderation decisions being a lens over the same data.