Quoting Michael Paoli (Michael.Paoli@cal.berkeley.edu):
Impacts should mostly be pretty minimal
Yay for DNS secondaries. (My domains have _four_ secondaries with diverse locations, server software, and management, in addition to ns1.linuxmafia.com.)
Yay for SMTP robustness. Downtime was about 2 days and 7 hours; SMTP retries are supposed to continue at least 4-5 days according to RFC 5321 section 4.5.4.1. So, I don't expect _any_ mail en-route to linuxmafia.com to fail, unless the sending MTA is extremely standards-incompliant.
It should be noted that the second robustness (SMTP retries) really works only if the first one also does (authoritative DNS fallback). Which is why, as a domain operator, one wants to have diverse secondaries and set DNS timeout values carefully.
Let's see how long the secondaries' zones would have continued to be valid if ns1.linuxmafia.com had remained offline (and I didn't otherwise fix the situation):
$ more linuxmafia.com.zone ;called as ORIGIN linuxmafia.com. $TTL 86400 @ IN SOA ns1.linuxmafia.com. rick.deirdre.net. ( 2018101400 ; serial 7200 ; refresh 2 hours 3600 ; retry 1 hour 2419200 ; expire 28 days 900 ; negative TTL 15 mins ) ; [...]
So, the DNS secondaries would have had me covered for almost a month (but SMTP mail would have started bouncing in a couple of days).