How to combine large number of key-value pair tables into a single table?

Question

I have 250+ key-value pair tables with the following characteristics
1) keys are unique within a table but may or may not be unique across tables
2) each table has about 2 million rows
What is the best way to create a single table with all the unique key-values from all these tables? The following two queries work till about 150+ tables
with
  t1 as ( select 1 as key, 'a1' as val from dual union all
          select 2 as key, 'a1' as val from dual union all
          select 3 as key, 'a2' as val from dual )
, t2 as ( select 2 as key, 'b1' as val from dual union all
          select 3 as key, 'b2' as val from dual union all
          select 4 as key, 'b3' as val from dual )
, t3 as ( select 1 as key, 'c1' as val from dual union all
          select 3 as key, 'c1' as val from dual union all
          select 5 as key, 'c2' as val from dual )
select coalesce(t1.key, t2.key, t3.key) as key
,      max(t1.val) as val1
,      max(t2.val) as val2
,      max(t3.val) as val3
from t1
full join t2 on ( t1.key = t2.key )
full join t3 on ( t2.key = t3.key )
group by coalesce(t1.key, t2.key, t3.key)
with
  master as ( select rownum as key from dual connect by level <= 5 )
, t1 as ( select 1 as key, 'a1' as val from dual union all
          select 2 as key, 'a1' as val from dual union all
          select 3 as key, 'a2' as val from dual )
, t2 as ( select 2 as key, 'b1' as val from dual union all
          select 3 as key, 'b2' as val from dual union all
          select 4 as key, 'b3' as val from dual )
, t3 as ( select 1 as key, 'c1' as val from dual union all
          select 3 as key, 'c1' as val from dual union all
          select 5 as key, 'c2' as val from dual )
select m.key as key
,      t1.val as val1
,      t2.val as val2
,      t3.val as val3
from master m
left join t1 on ( t1.key = m.key )
left join t2 on ( t2.key = m.key )
left join t3 on ( t3.key = m.key )
/

Answer 1

A couple of questions, then a possible solution.
Why on earth do you have 250+ key-value pair tables?
Why on earth do you want to consolodate them into one table with one row per key?
You could do a pivot of all of the tables, without joining. something like:
with
  t1 as ( select 1 as key, 'a1' as val from dual union all
          select 2 as key, 'a1' as val from dual union all
          select 3 as key, 'a2' as val from dual )
, t2 as ( select 2 as key, 'b1' as val from dual union all
          select 3 as key, 'b2' as val from dual union all
          select 4 as key, 'b3' as val from dual )
, t3 as ( select 1 as key, 'c1' as val from dual union all
          select 3 as key, 'c1' as val from dual union all
          select 5 as key, 'c2' as val from dual )
select key, max(t1val), max(t2val), max(t3val)
FROM (select key, val t1val, null t2val, null t3val
      from t1
      union all
      select key, null, val, null
      from t2
      union all
      select key, null, null, val
      from t3)
group by keyIf you can do this in a single query, unioning all 250+ tables, then you do not need to worry about chaining or migration. It might be necessary to do it in a couple of passes, depending on the resources available on your server. If so, I would be inclined to create the table first, with a larger than normal percent free, then do the first set as a straight insert, and the remaining pass or passes as a merge.
Another alternative might be to use the approach above, but limit the range of keys in each pass. So pass one would have a predicate like where key between 1 and 10 in each branch of the union, pass 2 would have key between 11 and 20 etc. That way everything would be straight inserts.
Having said all that, I go back to my second question above, why on earth do you want/need to do this? What is the business requirement you are trying to solve. There might be a much better way to meet the requirement.
John