fault-tolerant CAS clustering for the impatient

[Nathan Schimke]

I've read through the list archives, and I've found some tantalizing
hints but no clear answer to this:

If I'm looking for a fault-tolerant CAS service for a multi-tier
application (fault-tolerant in the sense that a server can "disappear"
and the end users aren't affected), am I really stuck with rolling my
own implementation or waiting for the CAS 3.1 release?

Or has somebody got an implementation which works in production and
they'd be willing to share, even if they can't actively support it?