But my given understanding is that a great many sites have
relatively very short TTLs set for their severs and
intranets, and that in many cases, even a matter of a
25 hrs, can cuse a gret many of them to expire.
Local ISPs can get away with this, and some do because they have under-powered DNS servers. The big upstreams (verizon uunet, bbn, etc) have much longer TTLs. Those networks are monitored 23/7. If a root server goes down they switch to their backup automatically and when they have less than two of three root servers available the first thing they do is terminate the ttl timeout. (This is automatic, nobody has to do anything to make this happen). This preserves their existing tables.
But as I mentioned, most of the big boys don't bother enforcing ttls at all anymore. If they get a replacement IP they honor it immediatly, but if the do not get any update they don't expire anything either. Its just not cost effective.
BTW, as I recall the longest outage of ALL root servers at the same time was 1 hour, and it was done as a test, after carefull coordination.