The test e-mail is an e-mail from the registrar to the contact address, so if it bounces, the register is then required to send a real e-mail to that address, which will presumably also bounce. Then the countdown starts.

I have very little expectation that the boneheads who came up with this scheme considered this possibility, and therefore exempted test e-mails from the bounce-trigger requirement.

The imaginary black-hat that hangs out on my left shoulder has already suggested that spoofing a bounce message for a correctly delivered registered e-mail could be used for mischief. Phishing the domain customers is entirely too obvious for him, though the imaginary white-hat that hangs out on my right shoulder seems quite concerned about it. They both agree that black-hat wins this round.