Supongo que no puedes simplemente:
select
from tbl_x
window w as (partition by ID order by Date)
where col_a is null;
Si no es así, es posible que deba:
select
Col_A,
Col_B,
case when Col_A is null
then (select col_a
from tbl_x x2
where x2.id = x1.id
and col_a is not null
and x2.date < x1.date
order by date desc
limit 1)
else Col_A
end Col_A_Lag,
case when Col_B is null
then (select col_b
from tbl_x x2
where x2.id = x1.id
and col_b is not null
and x2.date < x1.date
order by date desc
limit 1)
else Col_B
end Col_B_Lag
from tbl_x x1;
Con una indexación adecuada, el rendimiento probablemente sería bastante bueno.
create index ... (id, date desc, col_b) where col_b is not null;