Chris,

> Have you run test_slony_state to check the state of the cluster?
>
>   http://slony.info/documentation/monitoring.html#TESTSLONYSTATE
>
> It's likely that confirmations aren't getting back to the origin,
> which tends to imply some form of misconfiguration or a network
> problem.
>
> That script should help give some idea as to where things are breaking
> down.

I've been able to re-create the problem in isolation, away from my
live cluster, using two virtual machines (One MS Windows master with
all slons (the slony service), running Hiroshi Saito's Slony 2.0.2
package, the other a Linux machine that I've built 2.0.2 on ). This is
perhaps the simplest possible replication scenario: one master, one
slave, no failover or cascading, with a slonik script more or less
copied verbatim from the docs' "replicating your first database"
example. To my mind, this suggests that the problem is not in fact
down to misconfiguration.

I'll run those tests tomorrow in work and report back what they find.

Regards,
Peter Geoghegan