On 3rd of April 2019, Alvaro Herrera committed patch:

Log all statements from a sample of transactions
 
This is useful to obtain a view of the different transaction types in an
application, regardless of the durations of the statements each runs.
 
Author: Adrien Nayrat

Commit message makes it pretty clear, so let's see this in action.

I'll need a sample table:

=$ create table test (
    id serial primary key,
    int_1 int4 not null default 0,
    int_2 int4 not null default 0
);
CREATE TABLE

Now, let's make some test transactions.

Let's assume that I will want transactions that insert a row, and then update it.

Something like:

=$ BEGIN;
=$ insert into test (int_1, int_2) values (random() * 100000, random() * 100000);
=$ update test set int_1 = int_1 - 50, int_2 = int_2 - 100 where id = currval('test_id_seq');
=$ COMMIT;

Nothing really fancy. All of these queries will be fast. Now, let's make sure we run quite a lot of them:

$ (
    echo "BEGIN;"
    echo "insert into test (int_1, int_2) values (random() * 100000, random() * 100000);"
    echo "update test set int_1 = int_1 - 50, int_2 = int_2 - 100 where id = currval('test_id_seq');"
    echo "COMMIT;"
) > single.sql
 
$ for i in {1..10000}; do cat single.sql ; done > 10k.sql

Now, running this via:

$ psql -f 10k.sql -qAtX

generated logfile with size of ~ 5.5 MB (with log_min_duration_statement = 0).

So, let's now change the settings:

=$ alter system set log_min_duration_statement = 1000;
ALTER SYSTEM
 
=$ alter system set log_transaction_sample_rate = 0.001;
ALTER SYSTEM
 
=$ select pg_reload_conf();
 pg_reload_conf
────────────────
 t
(1 row)

And now, redoing the 10k.sql file generated only ~ 6.5kB of LOG! With 11 transactions in it, so pretty close to expected 0.1%.

This is great. Thanks, Adrien and Alvaro.