The idea is that if you are dependant on a single switch for critical services e.g. Single attached servers or critical users (say dealers in the financial industry, the boss) then conside Dual SE.
For a network core/distribution layer it is possible to have resilient paths and there is no need in that case for Dual SE. The network keeps working even if a whole box goes down. This can also apply where there are dual connected servers.
AH! You say. Don't I gert extra resilience if I just add dual SE everywhere.
No.
Since the extra complexity means:-
There is more to go wrong so failure is more likely. There are more failure modes to evaluate and test. (and it is hard enough to do this in even the simplest cases) Probably more good reasons here but it is nearly time to go to the pub.
Check out posts form Vincent in this forum for a lot of good material. Buy his book too! Look at his papers on his web site. It has a WAN focus IIRC but many of the principles apply everywhere.
I think that cisco have material on their web site that has similar statements to the above.
Look for SRND (Solution Reference Network Design) s.