How to create an SCD2 type table from partitioned table PostgreSQL/Clickhouse?

Your idea combining window functions with aggregation, could work:

1. Widen your frame to between unbounded preceding andunbounded following to override the default between unbounded preceding andcurrent row that comes with an order by and no frame definition.
2. In each row, compare all columns to lag() of all columns. Use is distinct from to make it null-safe. Add a failover self-comparison for the first row which otherwise wouldn’t have any lag() to compare to.
3. Wrap that in a subquery and use a stepping/tumbling count(*) to establish sequences of duplicates.
4. For each sequence of duplicates for a given user_id and country_id, take its min(ds) as dt_start and max(ds) as dt_end.
5. Use not bool_or(is_latest) or bool_and(not is_latest) to see if the group contains the last record, and if so, swap dt_end for a null.

_{demo at db<>fiddle}

select min(ds) as dt_start 
     , case when every(not is_latest) then max(ds) end as dt_end
     , user_id, country_id, n_sessions_1d, n_sessions_3d, n_sessions_1w, n_sessions_2w, n_sessions_1m, total_time_spent_1d, total_time_spent_3d, total_time_spent_1w, total_time_spent_2w, total_time_spent_1m, is_subscription_1d, is_subscription_3d
from(select*, count(*)filter(where is_changing)over w2 as sequence_of_duplicates
     from(select*, (n_sessions_1d, n_sessions_3d, n_sessions_1w, n_sessions_2w, n_sessions_1m, total_time_spent_1d, total_time_spent_3d, total_time_spent_1w, total_time_spent_2w, total_time_spent_1m, is_subscription_1d, is_subscription_3d)
                   is distinct from
                   lag((n_sessions_1d, n_sessions_3d, n_sessions_1w, n_sessions_2w, n_sessions_1m, total_time_spent_1d, total_time_spent_3d, total_time_spent_1w, total_time_spent_2w, total_time_spent_1m, is_subscription_1d, is_subscription_3d)
                        ,1,(n_sessions_1d, n_sessions_3d, n_sessions_1w, n_sessions_2w, n_sessions_1m, total_time_spent_1d, total_time_spent_3d, total_time_spent_1w, total_time_spent_2w, total_time_spent_1m, is_subscription_1d, is_subscription_3d))
                   over w1 as is_changing
                 , lead(ds)over w1 is null as is_latest
          from public.app
          window w1 as (partition by user_id, country_id
                        order by ds
                        rows between unbounded preceding 
                                 and unbounded following) )s1
     window w2 as (partition by user_id, country_id
                   order by ds
                   rows between unbounded preceding 
                            and current row) )s2
group by sequence_of_duplicates, user_id, country_id, n_sessions_1d, n_sessions_3d, n_sessions_1w, n_sessions_2w, n_sessions_1m, total_time_spent_1d, total_time_spent_3d, total_time_spent_1w, total_time_spent_2w, total_time_spent_1m, is_subscription_1d, is_subscription_3d
order by 1, user_id, country_id;