BUG #6607: Strange select behavior

The following bug has been logged on the website:

Bug reference:      6607
Logged by:          Sheplyakov Alexander
Email address:      suvisor.root@gmail.com
PostgreSQL version: 9.1.3
Operating system:   Debian GNU/Linux Wheezy kernel ver 3.3.1
Description:=20=20=20=20=20=20=20=20

Create test table in such manner:
create table testt (id serial PRIMARY KEY, val text not null);
Populate table with test data:
 insert into testt (val) values ('1'), ('2'), ('3'), ('4'), ('5'), ('6'),
('7'), ('8'), ('9');
(repeat next line until table doesn't contain approx 1000000 rows)
insert into testt (id, val) select nextval('testt_id_seq'::regclass),
currval('testt_id_seq'::regclass) + val::integer from testt;

And then, repeat some times such query:
select * from testt where id =3D (random()* 100000)::integer;

And sometimes it comes out something like this:
id | val
-------+--------
11894 | 15051
29233 | 42198
80725 | 90213
85688 | 100992
88017 | 108075
(5 rows)
Here can be 2, 3 or other rows amount in result... But must be only one!