I've edited to the title to be just about total capacity, because that appears to be what the question is about.
Total capacity represents the rear derailleur's ability to keep the chain slack (prevent drooping) down at the small/small gear combination even when the chain is properly sized long enough to accomodate the large/large combination.
The definition of total capacity is universally accepted in the design sense, but the implications of it are not, especially if you look across history. There's a lot of liability-consciousness built into the current manufacturer-endorsed premise that everyone must always have a bike set up to make it impossible to shift into a combination outside the RD total capacity, especially on the slack chain end.
Part of the question is how important is it to never find oneself in a small chainring/small-end-of-the-cluster gear combination where the chain goes slack. To answer that, one could point to the various early derailleurs where riding with the chain slack was an inevitability, and also to the period in the history of cycling where riding with the chain under spring tension was considered a potential major efficiency loss. Looking at those examples as evidence that it's a good idea to just not care about having the chain droop is bad. Untensioned chains can derail and jam, and it's not totally predictable when and why. Mostly it happens in aggressive riding situations, but exceptions are possible. However, it does bear pointing out that for the most part it does not happen in low-speed or low-intensity moments.
It is true and has always been true that some very high-mileage, prolific cyclists are animals on the bike and never develop a great intrinsic of what the components are doing and how to treat them. But, that is not most of us. Most regular users of derailleur bicycles always have been able to get familiar enough with how the systems work that staying out of small/small combinations is intuitive and easy if we want to stay out of them.
Cyclists in the latter camp, those with good control and intrinsic feel for the gear they're in, can basically cheat total capacity on the slack chain end as much as they want. You can look at a list of your cog sizes and make the decision for yourself where in the range you know you're not going to find it useful to shift further into the small/small combinations for your riding style. You can also have your own sense of what the situation is likely to be if you do shift into a slack chain combination by accident. There's a big difference in this regard between a road cyclist with a triple who is only really using the small chainring as a bailout gear and a mountain biker who will need to stay in it throughout technical climbing sections where they're throwing the rear end of the bike around.
The chain should literally always be set up long enough to accomodate large/large (with max chain growth if applicable) on everyone's bike without destroying anything. There's just no good reason to cheat it in that direction. The one exception is if you break a chain while out on a ride and then excise the damaged part to finish it out and get home. In that case the answer for most cyclists is just be careful to stay out of the gear combinations where the rear derailleur is over-stretched. If it's a child or inexperienced cyclist that got the broken chain, using the limit screws to lock those gear combinations out completely is reasonable.