normalise data across duplicates but keep field containing overwritten values-possible?

I have a file of company name exactly duplicated. Within each duplicate set of rows each row can have a different account code (number).

I need to normalise all rows to one account code (any one will do - I can actually do this bit using some commercial software so this bit can be skipped as part of this thread) but also keep every variation/different account code in an 'audit trail' field for each row.

The attached sample file shows the input and desired output i would like.

Any suggestions gratefully received!

Dave