One successful replay proves only the first starting state. Repetition tests whether the workflow resets correctly and timing remains stable.
Why one test is not enough
There is no universal number, but at least three consecutive clean cycles on realistic test data is a sensible minimum for a low-risk task.
Use a staged rehearsal
Run one observed cycle, then three consecutive cycles, then a limited real batch in MacroLoop. Inspect the output after every stage and record the stop condition.
- Single observed replay
- Three-cycle repeat test
- Limited real batch
- Independent result check
Risk matters more than count
Even many successful tests do not justify unattended high-consequence work. Interface updates and unexpected data can still break the path.
THE TAKEAWAY
Use staged evidence and keep the first unattended batch small.