The tragedy of the commons is a term used in to describe a situation in a shared-resource system where individual users acting independently according to their own self-interest behave contrary to the common good of all users by depleting or spoiling that resource through their collective action.
There are a few fallacies in this term when using it as an argument/problem for SafeNet and its reliability. @Warren touches on this further up in this thread. The term has a very narrow scope taking very destructive intentions into consideration assuming all factors are known. Lets entertain the discussion anyway. It could possibly apply to Blockchain technology as all data is shared (shared resource) and propagation delay in regards to scalability will be affected and may even halt the network at some point. But lets focus on Maidsafe.
Resources in Safenet are not shared with everyone, everyone has equal limited access, only whats needed. I guess it will be clearer when the paper on disjoint groups are released (sharding). I personally look forward to understand better myself. Anywho, a piece of data that Alice has stored with Bob doesnt need to be shared with Eve, it will be stored at enough places to be retreivable with privacy and randomness so it cannot be targeted in any way. This also removes the attack vector of shutting down/overload disjoint groups. I like Davids comparison to ants where no single entity needs to know everything, yet work in complete unison for the better good of the network. Many Alices need to store a great amount of data to even have an effect but I will get to that further down.
As @Antifragile touches upon further up this thread, every time data is stored someone is rewarded, this reward will have certain value. Its important to keep in mind that the network is autonomous, not a human controlled system (hence the requirement of certain level of completion before Safenet can launch) where individual users act independently according to their own self-interest and contrary to the common good. In Safenet, every individuals self interest is ALIGNED with common (network) good. If you think you can bloat the system, all you do is reward other participants which increases the value of the network and attracts more participants. The network will (automatically) adjust the rewards and if resources are scarce there will be an arbitrage opportunity which the free market exploits (free market assumption is based on fungibility). The only way this logic would fail is if people suddenly lose interest in wealth, which is beyond the philosophical scope of this discussion.
Another point that @fred mentions further up is the redundancy of Moores law, does not apply to storage whatsoever. For somebody to attempt such an attack where they cause an immense amount of PUTs to then immediately go offline to never return will cost the attacker much more than the network, if the attack is targeted to a “popular” piece of data (read DDoS) the network only gets stronger and faster as every participant enforces the data. The supply of storage and accessibility due to privacy will make farmers very wealthy as the demand for both the coins and securing the network (farming) increases under such an attack (Need coins to create PUTs and storage providers will be needed, thus higher rewards if scarce). If the network was not autonomous this could be exploited through forced inflation with sudden cut of demand, such instance is taken care of by the network.
I personally dont see this as a valid issue/threat.