Those of you paying attention noticed that I just released 1.2.23 and 
2.1.1 but not 2.0.8

We had applied the fix for bug #255 the 'serialization conflicts with 
PostgreSQL 9.1' on the 2.0 branch in addition to the 2.1 branch.

When testing 2.0.8 RC2 I noticed that once in a while the disorder - 
"MoveSet" test in the clustertest framework gave me failures where the 
nodes could end up with different data.   The MoveSet test does 11 move 
sets between nodes in the cluster.  If I run the MoveSet test in a loop 
10 times I would typicaly get one failed test run.

On the version before the bug #255 commit I was never able to reproduce 
that failure on many runs.

I have also not been able to reproduce that failure on the 2.1.1 with 
which includes a fix for #255. (which is why I released 2.1.1)

(All of the above is with PostgreSQL 9.1).

My inclination is to take out the fix for #255 from 2.0.8 and also put 
back in the warning about 9.1 "might be unsupported" that you would see 
with 2.0.7.

We can also leave the bug #255 fix in and put the warning back.

If someone can figure out why this is failing in 2.0 then we do talk 
about addressing that.  I can't rule out it being a problem in the test 
itself, but nothing pops out at me  (A version of the disorder tests 
back-ported to 2.0 can be found at 
https://github.com/ssinger/slony1-engine/tree/REL_2_0_STABLE-clustertest)