We are having strange issues with the SQL Server bulk load stage after 7 upgrade. We have several jobs that we used in a previous POC that worked correctly. These are pretty simple jobs that pull data from an EBCDIC formatted files, run them through a set of parallel conversions (using the splitter/combiner) then bulk loading them into SQL Server.
While testing the jobs I found out that we were getting duplicates in the primary key.
I could not understand what was going on until I removed the primary key constraints from the table and saw that every row that was being inserted was all zeros. 
I wanted to make sure it was nothing in the transformation so I ran the code in the debugger then finally I redirected it to a flat file and the numbers were OK. 
Once I did that I decided to modify the original job to load both into a transformer so that I could write to the file and the table at the same time, but once I did that everything was OK with the data going into SQL Server. This made me think that when I refreshed the columns that the metadata was updated and it worked because of that.
Finally, after a little more experimenting it looks like I need to have a transformer stage after a combiner and before going into the bulk load or it loads zeros.