By which columns to filter data that has already been loaded and load only new data?

Question

SQL

I need to implement incremental data loading.

example:
A piece of the table looks like this

select * from order_states limit 20

+--------+-------------+-------------------+-------+-------------------+------+-----------+--------+-------+------------+
|order_id|ext_source_id|                 dt|  state|                 ts|msk_dt|api_user_id|state_id|comment|ext_state_id|
+--------+-------------+-------------------+-------+-------------------+------+-----------+--------+-------+------------+
|     919|            1|2021-01-29 16:48:27|   PROC|2021-02-12 00:12:20|  null|       null|       7|   null|          72|
|     920|            1|2021-01-29 16:48:30|PRINTED|2021-02-12 00:12:20|  null|       null|       1|   null|          73|
|     920|            1|2021-01-29 16:48:30|   DONE|2021-02-12 00:12:20|  null|       null|       1|   null|          74|
|     921|            1|2021-01-29 14:48:30|   SAVE|2021-02-12 00:12:20|  null|       null|       6|   null|          75|
|     921|            1|2021-01-29 14:48:30|   PROC|2021-02-12 00:12:20|  null|       null|       7|   null|          72|
|     922|            1|2021-01-29 14:48:00|    NEW|2021-02-12 00:12:20|  null|       null|       9|   null|          76|
|     923|            1|2021-01-29 14:48:31|PRINTED|2021-02-12 00:12:20|  null|       null|       1|   null|          73|
|     923|            1|2021-01-29 14:48:31|   DONE|2021-02-12 00:12:20|  null|       null|       1|   null|          74|
|     924|            1|2021-01-29 14:48:21|   SAVE|2021-02-12 00:12:20|  null|       null|       6|   null|          75|
|     924|            1|2021-01-29 14:48:22|   PROC|2021-02-12 00:12:20|  null|       null|       7|   null|          72|
|     925|            1|2021-01-29 14:48:52|PRINTED|2021-02-12 00:12:20|  null|       null|       1|   null|          73|
|     926|            1|2021-01-29 14:48:04|    NEW|2021-02-12 00:12:20|  null|       null|       9|   null|          76|
|     927|            1|2021-01-29 14:48:23|   SAVE|2021-02-12 00:12:20|  null|       null|       6|   null|          75|
|     925|            1|2021-01-29 14:48:52|   DONE|2021-02-12 00:12:20|  null|       null|       1|   null|          74|
|     928|            1|2021-01-29 14:48:04|    NEW|2021-02-12 00:12:20|  null|       null|       9|   null|          76|
|     927|            1|2021-01-29 14:48:24|   PROC|2021-02-12 00:12:20|  null|       null|       7|   null|          72|
|     929|            1|2021-01-29 13:48:39|PRINTED|2021-02-12 00:12:20|  null|       null|       1|   null|          73|
|     930|            1|2021-01-29 14:48:24|PRINTED|2021-02-12 00:12:20|  null|       null|       1|   null|          73|
|     929|            1|2021-01-29 13:48:39|   DONE|2021-02-12 00:12:20|  null|       null|       1|   null|          74|
|     931|            1|2021-01-29 14:48:04|PRINTED|2021-02-12 00:12:20|  null|       null|       1|   null|          73|
+--------+-------------+-------------------+-------+-------------------+------+-----------+--------+-------+------------+

as a waterpoint, I first chose the order_id column
and created a separate table where I stored the value of the waterpoint and displayed the missing data in this way:

select * from order_states where order_id > {значение вотерпоинта, например 1040}

But it seems to me that, judging by the data, this is not the most optimal way to filter the missing values.
Which columns are better to filter the missing values ​​so as not to lose any data during additional loading? By dt and order_id at the same time?