That's a really interesting read (and worth much more attention than the pithy one-liners of people who just want to read the title).

On reflection, I think my take away is that Bluesky will always by necessity of its design be hosted and controlled by a single centralised company. But what their architectural model does allow is the possibility of a wholesale migration from one centralised provider to another. That is, it would be possible for a suitably resourced and motivated company to host its own mirror Relay and other components and have essentially a fully functional Bluesky clone. In the event that Bluesky ever "does a Twitter" and go into terminal decline, in theory this might mean that a successor/competitor could emerge and take on the network without loss of existing content.

I'm not sure that'll ever actually happen, but it's an interesting thought.