A couple of the big reasons I’ve seen disaster recovery plans fail are:
1) Insufficient “dress rehearsals.” Key to any disaster recovery plan is testing it in minute detail, and documenting all those “off books” steps that are required to get your environment restored. Many of the “oh, I’d always remember to do X” steps that seem so simple during a dry run are forgotten during a real disaster and need to be captured. Nothing should reside “in my head"”.
2) Design your plan so that it is “componentized” as much as possible, since you will never be able to account and plan for every type of disaster. What happens if communications are down? What happens if key leaders, managers or “doers” are unreachable? What happens if there is a more pressing priority than restoring IT infrastructure? What do you do if physical security is no longer a given for people or facilities? You’ll go crazy trying to map out every conceivable disaster scenario, but if you have plans for various people, processes and infrastructure failing, you can pick the right combination of plans based on a disaster that was completely unforeseen.