Having seen that all previous problems went unresolved, heres a bit
more info. The system is 32 bit, running on enterprise redhat 4.7. It
is slony's slave node, so it will be hit with quite few updates.
My guess is that it happened when we ere adding/removing slony to the
system for Nth time (due to it sometimes going out of sync).
Another thing that makes me think so, is what I've seen in pg_dump's output:
CREATE TRIGGER _simreplic_denyaccess_208
BEFORE INSERT OR DELETE OR UPDATE ON some_table
FOR EACH ROW
EXECUTE PROCEDURE 28799('_somereplic');
Which is obviously wrong.
hth.