> but nothing short of dropping and uninstalling node 1 (k) and
> reinstalling/subscribing gets it replicating again.
>
> The problem is repeatable after an init cluster.
>
> What happened please?

The problem appears to be a mismatch of slon versions.

The "latest" slon knows about the ACCEPT_SET event (which is a new event
created to prevent data loss when a MOVE_SET takes place; a node holds off
on applying changes until it has gotten both...)

But evidently one of the slons you are running doesn't include ACCEPT_SET
in its set of events.

Hmmph.  We haven't had code to make sure that there is an exact interlock
between versions in the "beta" editions; where 1.0.5 introduced the notion
of refusing to run if it found the DB was configured for some other
version, we haven't had anything formally incrementing to force the same
refusal to be true between different beta editions.

That's something the developers should probably think about a bit...