Gitlab's praefect enormous wal generation on postgresql

We have praefect's database about 500 Mb in size, but wal generation per week is relly huge - about 1Tb. Most frequent query is being executed ~8 times/minute:

WITH lock AS (
	SELECT
		id
	FROM
		replication_queue_lock
	WHERE
		id LIKE $1 || $4 || $2 || $5 AND
		NOT acquired
	FOR UPDATE SKIP LOCKED
)
, candidate AS (
	SELECT
		id
	FROM
		replication_queue
	WHERE
		id IN (
			SELECT DISTINCT
				first_value(queue.id) OVER(PARTITION BY lock_id, job ->> $6 ORDER BY queue.created_at)
			FROM
				replication_queue queue
			JOIN
				lock
					ON queue.lock_id = lock.id
			WHERE
				queue.state IN (
					$7
				,	$8
				) AND
				NOT EXISTS(
					SELECT
						$9
					FROM
						replication_queue_job_lock
					WHERE
						lock_id = queue.lock_id
				)
		)
	ORDER BY
		created_at
	LIMIT $3
	FOR UPDATE
)
, job AS (
	UPDATE
		replication_queue queue
	SET
		attempt = CASE
			WHEN job ->> $10 = $11 THEN
				queue.attempt
			ELSE
				queue.attempt - $12
		END
	,	state = $13
	,	updated_at = timezone(
			$14
		,	now()
		)
	FROM
		candidate
	WHERE
		queue.id = candidate.id
	RETURNING
		queue.id
	,	queue.state
	,	queue.created_at
	,	queue.updated_at
	,	queue.lock_id
	,	queue.attempt
	,	queue.job
	,	queue.meta
)
, track_job_lock AS (
	INSERT INTO replication_queue_job_lock(
		job_id
	,	lock_id
	,	triggered_at
	)
	SELECT
		job.id
	,	job.lock_id
	,	timezone(
			$15
		,	now()
		)
	FROM
		job
	RETURNING
		lock_id
)
, acquire_lock AS (
	UPDATE
		replication_queue_lock lock
	SET
		acquired = $16
	FROM
		track_job_lock tracked
	WHERE
		lock.id = tracked.lock_id
)
SELECT
	id
,	state
,	created_at
,	updated_at
,	lock_id
,	attempt
,	job
,	meta
FROM
	job
ORDER BY
	id;

Is there any way to tune application to slow this thing down?