That goes in the "that depends" bucket.
If the column is nullable, we are going to run effectively this as a check first
select A.rowid from MYTAB A where( "COL" is null)
which will read the table.
If the column is already not null, then we'll do something like
select A.rowid
from MYTAB A,
(select "COL" from MYTAB A
where( "COL" is not null)
group by "COL" having count(1) > 1) B
where( "A"."COL" = "B"."COL")
which can be done with an index FFS