Ниже для BigQuery Standard SQL
#standardSQL
SELECT DISTINCT user_id, created_at
FROM (
SELECT user_id,
ARRAY_CONCAT_AGG([created_at_a, created_at_b]) created_at_ab
FROM `project.dataset.table`
GROUP BY user_id
), UNNEST(created_at_ab) created_at
Вы можете протестировать, поиграть с этим, используя примеры данных из вашего вопроса, как показано ниже
#standardSQL
WITH `project.dataset.table` AS (
SELECT 1 id, 1 user_id, TIMESTAMP '2019-01-24 12:20:00 UTC' created_at_a, TIMESTAMP '2019-01-25 01:04:00 UTC' created_at_b UNION ALL
SELECT 2, 1, '2019-01-24 12:20:00 UTC', '2019-01-25 01:03:00 UTC' UNION ALL
SELECT 3, 1, '2019-01-24 12:22:00 UTC', '2019-01-25 01:03:00 UTC' UNION ALL
SELECT 4, 1, '2019-01-24 12:22:00 UTC', '2019-01-25 01:04:00 UTC' UNION ALL
SELECT 5, 2, '2019-01-24 20:48:00 UTC', '2019-01-24 20:49:00 UTC' UNION ALL
SELECT 6, 2, '2019-01-24 11:21:00 UTC', '2019-01-24 20:49:00 UTC'
)
SELECT DISTINCT user_id, created_at
FROM (
SELECT user_id,
ARRAY_CONCAT_AGG([created_at_a, created_at_b]) created_at_ab
FROM `project.dataset.table`
GROUP BY user_id
), UNNEST(created_at_ab) created_at
-- ORDER BY user_id, created_at
с результатом
Row user_id created_at
1 1 2019-01-24 12:20:00 UTC
2 1 2019-01-24 12:22:00 UTC
3 1 2019-01-25 01:03:00 UTC
4 1 2019-01-25 01:04:00 UTC
5 2 2019-01-24 11:21:00 UTC
6 2 2019-01-24 20:48:00 UTC
7 2 2019-01-24 20:49:00 UTC