"PostgreSQL Bugs List" <pgsql-bugs@postgresql.org> writes:
> Description: listeners not cleaned on crash
This is a deficiency, without doubt, but it's always seemed like a
mighty low-priority one to me. There already is a provision for
deleting entries when they are found not to refer to any existing
process, so the odds of actually sending a mistaken notify seem low.
> This is a very annoying bug if you are running slony1 replication system.
If slony reacts badly to an extra notification then I think that's a
slony problem more than anything else. It shouldn't be that fragile.
regards, tom lane