You want to restart a result on your system because it errored out. Most likely too late. The result file, good or bad is already uploaded by the time you notice. If an error occurs and it's recoverable, the task is already automatically restarted from either the beginning or from last checkpoint.

Just wonder, how many do you have like that to want that function? Suggest that if you modify the core OS system, to include the clock, something BOINC under circumstances thoroughly dislikes, you stop/exit the client first. You can even automate that by adding <exclusive_app> lines to the BOINC configuration.