So you want to:

- Convert the comments to rows
- For each of these generated rows, link to one of the existing table rows
- If there's more rows from the comments than in the table, fill down the IDs

?

If so, you can do something like:

- Split the comments to rows; assign a row number to each distinct value using dense_rank()
- Assign a row_number to each row in the table
- Partition outer join by ID the table to the generated rows. This will fill down the IDs for you:

Create table testdrop(ID Number,Name Varchar2(10), Comments Varchar2(100));

insert into testdrop values (1,'ABC','Test1; Test2:Test3');
insert into testdrop values (1,'XYZ','Test1; Test2:Test3');

with rws as (
select 
  distinct
  trim(regexp_substr(t.comments, '[^;:]+', 1, levels.column_value))  as error, 
  dense_rank() over (order by trim(regexp_substr(t.comments, '[^;:]+', 1, levels.column_value))) rn
from 
  testdrop t,
  table(cast(multiset(select level from dual connect by  level <= length (regexp_replace(t.comments, '[^;:]+'))  + 1) as sys.OdciNumberList)) levels
), t as (
  select t.*, row_number() over (order by id, name) rn from testdrop t
)
  select id, name, error from rws
  left join t partition by (id)
  on     rws.rn = t.rn
  order  by rws.rn;

ID  NAME  ERROR  
1   ABC   Test1  
1   XYZ   Test2  
1         Test3

For more on partitioned outer joins, read this by Adrian Billington:

http://www.oracle-developer.net/display.php?id=312