First I'll talk about redundancy. It's a specific measure of how well the list mitigates the results of failure. I consider a unit to fail when it doesn't live up to it's expected potential. So if a unit is either destroyed before performing its intended role, or isn't effective when it does perform its role, it fails. Redundancy is a good way around this. You take multiple reliable and expendable options to fulfill a role. This way you have backup in the case of a failure.
When optimizing to redundancy you build units on a budget. You don't waste points on upgrades you don't need. You build highly specialized units, though units that can fulfill a secondary role on the cheap are highly prized.
The opposing measure of list optimization that is fueling the current debate is reliability. When you optimize to reliability you try to make units more potent, so that when they fulfill their roll, they do it much better than required on average. You're try to reinforce a unit so that it is less likely to survive to fulfill its role and to effectively perform that role when needed.
If you are optimizing for reliability, you end up paying a few extra points to up squad sizes or purchase upgrades that are effective but useful only 10% of the time. Generalist utility units are very important.
Both methods of optimization have pit-falls associated with them. Go for too much redundancy and you've pulled the teeth out of your army. Everything is so expendable that it has no tactical value or offensive punch. MSU vanilla marines can really fall into this trap. 5 man squads in Razorbacks are potent in mass, but can't do much on their own. Same goes for many other MSU favorites.
Optimize for reliability too much and you put all your eggs in one basket. You can mitigate chance, but you can't control it completely. As many points as you put into making a unit reliable, you can still suffer the effects of a few bad rolls and loose a much larger investment in points. Fateweaver is my favorite example of this. He makes the demon army very reliable, but a few bad rolls and the demon player's army is crippled.
It's a give and take situation with reliability and redundancy. Make a unit too reliable and there aren't enough points left over to take a redundant back up. Build in too much redundancy and you don't have any clutch unit to fight the army out of a tough spot. I think wolves are such a good army because they can balance redundancy and reliability at the same time. They can add a wolf guard and special weapon to small units to make them reliable. And they have reliable units like thunder wolves that are not too expensive to preclude redundancy. Compare that with vanilla marines. TH/SS terminators are very reliable, but are so expensive when you factor in their transport requirement that you have to build a list around them. 5 man tactical squads in razors make a good redundant core, but you can't improve their reliability until you go to 10 man squads.
The best armies and units are the ones that can best balance reliability and redundancy. I think that a lot of armies are not getting the fair evaluation they should because players are applying the wrong optimization approach when building their lists. People are going too far down the MSU route for Dark Eldar and can't build the reliable Tyranid Nidzilla they used too. I think the xenos books need the right balance to work properly more than the imperial armies. Any of the marines can work if you get the balance wrong because they can fall back on their good stat line. Guard can work out of balance because you can gain reliability though sheer weight of firepower.