> This is the kind of system failure that we need really good and well tested disaster recovery plans for.
All the cool kids offer their services over multiple TLDs and have their name servers of record in multiple TLDs, too. It's not quite best practices for recursive DNS to regularly fetch the complete root zone to cache it, but it's not unreasonable to do so.
Having nameservers on multiple TLDs is definitely smart. But shouldn't glue address this at least for short outages already? How does the delegating zone handle and cache glue records?
Providing the same site on multiple TLDs is kind of a phishing hazard though. It may make sense for non end user facing APIs and other critical points, but for end user sites I think sites should stick to one domain. Otherwise they train users to fall for phishing sites.
All the cool kids offer their services over multiple TLDs and have their name servers of record in multiple TLDs, too. It's not quite best practices for recursive DNS to regularly fetch the complete root zone to cache it, but it's not unreasonable to do so.