Dealing With Error Records - Best Practice Question

There's lots of smart people here who can, I am sure, offer insight on this one.

I need to add handling of erroneous data to my package.

This is what my SSIS package does. The source table has thousands of records in it, most of which have been processed already (have a status of "Finished".) A typical run would have 400-600 records to process. The runs happen overnight.

Steps:
1) Execute SQL task that Updates all records in input table with status "New" or "Error" to "Process"
2) Data Flow task that takes as input all records with status "Process" and outputs to destinate OLEDB table
3) Execute SQL task that Updates all records with status "Process" to "Finished"


Erroneous records are identified at step 2 (Data Flow task) and need to be marked as "Error" so that they are not marked as "Finished" by step 3. They will be picked up and reprocessed by step 1 next time around.

I am having trouble seeing what would be the best way to achieve this? I am concerned that any steps I take might be prone to deadlocking as the updated data is also the source data?

My initial thought is to output some data to an "error" table in step 2, and then insert another Execute SQL task before step 3 that will update the source table to mark all these records as "Error" records. Does this make sense? Is there a best-practice way to achieve this?

Dylan.