Another option for cost savings is on the fly demand pricing of meals. Through some process of ingredient availability, customer analytics, registered interest(voting or something), etc, the centralized kitchen could decide to make a particular dish at a greater volume/lower cost which could then be passed on to consumers. i.e. Tuesday is 'spaghetti night' and you can feed your family for 50% less than other options.
I believe this may be how many cafeterias operate. Cafeterias are a good analogy because you have an environment already saving on wait staff and gratuities.
This might seem odd to a typical HN reader who spends $300+ on sushi dinners in SF, but this sort of thing could appeal to many people with less income and without the time to prepare meals themselves. It probably won't work until we can automate the delivery process, but I think at some point the economics will work out in certain markets.
There's nothing inherent to the delivery app business that makes this possible. This has always been possible, but it doesn't really happen on any significant scale.
This is happening right now: see 'ghost kitchens' on delivery apps, with three or four different 'restaurants' that are actually operating out of the same location.