The cost. There will be reasons to try and sabotage the network, but flooding it with data is both costly and ineffective. The network doesn’t care, it doesn’t know the value of data. Data that is paid for but never accessed it ends up subsidising data that is accessed, making the network more attractive to farmers and users, more robust.
If you want to harm the network there will be more effective and cheaper ways so I don’t believe there’s much risk of the above happening, nor of it causing a problem if it does.
If you are correct that in the end some data will need to be dropped, then the design could be amended, so IMO it’s not worth worrying about this at this stage. There are too many unknowns to know how all this will pan out so we must run experiments. So why not aim high? Trying for a perpetual network is surely a worthwhile experiment even if you don’t expect it to work.
I take the same attitude when people criticise Pay the Producer (PtP) on economic grounds, with IMO ideas based in the past rather than a vision for something original, which can change the rules of the game for the better. I can imagine benefits and I want us to see what they might be. It can always be disabled it if it is found not to further the goals we signed up for (the Safe Network fundamentals).