if using smart contracts spawns memory leaks then we have a much larger problem than node bloat 
They probably don’t. However, if you feel that gastoken is a useful economical tool (store gas when blockchain is idle, release gas and thus disk space when the blockchain is busy) then you can propose an ECIP that would integrate gastoken into EVM via a precompiled smart contract that simulates current behavior without actual disk bloat.
I’m not going to be the one to do the math - you are the one coming up with these memory claims. I understand where you are coming from and appreciate you caring about the ecosystem, but throwing numbers like these needs to have factual backing. How much disk space does 1 GST2 occupy? It’s definitely on the order of kilobytes (at best) rather than gigabytes. Now calculate how much $ETC it costs to mint 1 GST2, and what GST2 total supply must be in order to get to 1 Tb of disk space. I’m sure you’ll find that understuffed teams will not have enough money to produce such bloat.
Current $ETH blockchain, which is widely used for all sorts of dumb stuff, currently takes 137Gb of disk space. Not even close to 1 Tb.
I think GasTokens are useful as they represent an options contract on blockchain’s utilization and allow regular blockchain users (i.e. oracles, trading bots) to have more predictable & less volatile costs - which is definitely a major plus for businesses that are trying to build on ETC.
P.S.
One can use filesystems which will do compression & deduplication for you without having to modify geth or parity code.
https://btrfs.wiki.kernel.org/index.php/Compression
https://pthree.org/2012/12/18/zfs-administration-part-xi-compression-and-deduplication/
Would love to see a benchmark of how helpful such techniques might be.