I'm trying to get some work done tonight, running slony 1.2.13 (will
gladly update if this issue has been addressed) and during an
operation:

select "_slony_www".storeSet(203, 'set to alter tables');

it is scanning every single attribute of every single table in my
database, whether it is under replication or not.  This command above
is generating sql like this:

SELECT a.attnum, a.attname, a.atttypmod, a.attstattarget,
a.attstorage, t.typstorage, a.attnotnull, a.atthasdef, a.attisdropped,
a.attislocal, pg_catalog.format_type(t.oid,a.atttypmod) as atttypname
from pg_catalog.pg_attribute a left join pg_catalog.pg_type t on
a.atttypid = t.oid where a.attrelid = '78818327'::pg_catalog.oid and
a.attnum > 0::pg_catalog.int2 order by a.attrelid, a.attnum

If I go look in pg_class for the relfilenode that matches, it's in a
whole other schema that's not involved in replication at all.  Each
query is only taking a few milliseconds.  But with 40k or more
relations in the db, and it seems like a handful of 1 to 5 millisecond
queries for each one, slony is simply taking hours to do something
simple, and there's NO REASON for it to be looking at these tables in
other schemas that are not related.