Not only is all the VC subsidization questionable, but Doordash, Caviar, Uber Eats, etc. all strike me as just the shallow veneer of a delivery service. It's just about the leakiest abstraction of any major product that I can think of.
You have the restaurant, which is physically optimized for in-person meals. Take-out on its own is already fairly clunky at many sit-down restaurants, there's no waiting area, pick up counter, no easy way to pick up at the curb, etc. Pickup is clearly not the focus of those physical spaces.
Then you have the integration between the app and the restaurant. This is the one thing about the entire flow that is actually integrated in a meaningful way, but still it's only at the single point in time of placing the order. If anything goes wrong, if there's a delay in preparing the food, or anything is sold out, or if they end up giving you the wrong person's order, there's always a runaround of whether you should talk to the delivery support or the restaurant itself. Seeing as at least something goes wrong with probably 1 in every 3 or 4 orders I do, this loss is a probably a really big deal for profitability as I am always getting refunded for something that was missing or went wrong or getting a $10 credit for an unexpected 3 hour delay.
And then you have the delivery handoff itself. I suspect this mostly works well in the suburbs, but living in an apartment building this is yet another opportunity for things to go wrong. I would prefer not to have to come downstairs outside the building to pick it up when I'm paying ~$20 for this delivery to my door, but very often the delivery people complain they can't find the building (it's clearly marked on google maps), or that there's no parking (I live in a very dense area, why are all these apps so car-centered in their delivery crew? Why can't I request someone on a bike that wouldn't have this problem?).
Basically, delivery could be done very efficiently, and profitably if we made some bigger changes. There should clearly be specific restaurants, maybe even some that have no seats at all and are just a pickup kitchen, that are optimized for efficient pick-up and marked as such. Cities need to change their street design to accomodate how many more deliveries are happening today, vs when these streets were designed many years ago (this is not just a problem for food, look how many rideshares drop people off in the middle of the road because there's no space to pull over to a curb, or how many Amazon or UPS trucks block streets and bike lanes because there are no dedicated 5-minute stopping zones).
It reminds me of Marc Andreeson's recent "It's time to build" essay. Our whole world is structured around how things used to work 40 years ago, and nothing significant can ever be changed. So we have to have these crazy, expensive hacks like Doordash to try to shoehorn in new services that people want, instead of making deeper adaptations as residents' needs and preferences change.