So if we had 4 purely random macbooks, and 3 of them failed, then that would be a meaningful indication that the failure rate is higher than 10%?

The issue is both small sample and lack of randomness.