There is an argumentative premise in this framing which implies that Solana is cheap because they are pricing state wrong.
Issues with Solana downtime aside, Solana actually does have “state rent” which is effectively an existential deposit for each piece of storage. It’s just very cheap. They used to have an actual rent payments scheme but it is now required to have deposits. The main difference is just an eager pruning of state at pre-defined epochs - deposits are locked implicitly, not explicitly.
They use a simple argument, possibly oversimplified but directionally correct - it seems inspired by Arweave and goes like this: the deposit is 2x the cost of the disk space it takes to store the data for a year today on “enough” nodes. Because disk space decreases in cost by about half every year, when you integrate out to infinity the area under the curve is 2 times the area from 0 to 1. This also implies that properly calibrated deposits should decrease by half every year.
Estimating that cost is hard and can’t be done algorithmically - what’s important is the cost of estimation errors in either direction. Underestimates mean that state grows faster, although not unbounded - effectively an economic subsidy on activity. Overestimates mean that the state grows slower - effectively an economic depressant on activity. We should bear in mind these systems are reflexive and current activity seeds future activity.