The collaboration was generally fine: the contractor was in touch, the main functionality was implemented.

At the same time, after the changes were made, some things worked unstably or worse than before. Discrepancies arose on these points - they were considered as separate tasks for payment, although they appeared to be a result of the changes made.

Also, the deadlines were significantly shifted from the initially agreed ones.

I recommend clearly defining the criteria for task completion at the start, the approach to testing (especially without staging), and what is included in corrections after submission.