Incremental model without updating changed records - sql

Question

I want to build a dbt model that returns the minimum timestamp for each user_id.
My source table is a versioning table with several billion rows. Hence, I want to build it as an incremental model.
In source data, version_id is unique and each user_id can have several records. Here is what it looks like:
with t1 (version_id, user_id, updated_at, is_activated) as (
select * from values
(1, 'a', '2022-12-01'::timestamp, true)
, (2, 'b', '2022-12-02'::timestamp, true)
, (2, 'b', '2022-12-03'::timestamp, false)
-- , (1, 'a', '2022-12-03'::timestamp, true)
)
select * from t1
When a user_id gets another row (e.g. 4th row) with a newer time, I don't want dbt to update the old row as the first row has the minimum timestamp.
How to ensure the new rows do not update all the old ones based on the min timestamp per user_id?
{{
config(
materialized='incremental',
unique_key='version_id'
)
}}
WITH filtered AS (
SELECT
version_id
, user_id
, updated_at
FROM {{ ref('versions') }}
WHERE is_activated = TRUE
{% if is_incremental() %}
AND updated_at > (SELECT MAX(this.updated_at) FROM {{ this }} AS this)
{% endif %}
QUALIFY ROW_NUMBER() OVER (PARTITION BY user_id ORDER BY updated_at) = 1
)
SELECT * FROM filtered