"I would go for the auto-configuring software, though it is not
[...snip...] were splitting the existing partition."

I think you have made the point by detailing your approach: it leads to
complicated code in several places.

To me, the hallmark of a great approach is that complex stuff can be
epxressed in little code. The unix v6 kernel was less than 10,000 sloc
and the code reads almost like a book most of the time.

> True, but in this case you do not need to implement a lot, just take
> Sequoia and run your cluster. You have dynamic addition of nodes
> there too - a database adapter is required to implement
> backup/restore service.

I thought sequioa was a write-everywhere solution too. How does it deal
with 1,000+ updates/second and staying consistent at the same time?