Interview question

A new rebooking flow should cut support calls and must not increase flight cancellations through accidental self-service cancels. The cancellation rate in control is 2.0% (hypothetical). The product owner defines "harmless" as no more than 0.2 percentage points higher. After a test with 100,000 users per arm the treatment cancellation rate is 0.05 points higher and the superiority test gives p = 0.42, so the report says "no significant difference in cancellations, safe to launch". Is that right? Design the analysis you would have run and say how many users it needed.