Any more details about this would be interesting for me.
I’ve written off Solana after looking at it seriously several times now, yet it continues to have the appearance of thriving. Fwiw my biggest issues in the past has been a combination of large data and resource requirement and vulnerability of archive/full history nodes to Google services shutoff and centralization with censorship possibilities.
also two thirds of transactions are validators voting for blocks.
also afaik solana finalization time is not sub second, but 10 seconds. intrestingly why misleading info about sub second still persist...
also solana cannot really execute as part of single transaction all needed. so from 1/3 of useful transactions, half is just account lookup table creation.
so it feels somebody tries hard to tell the story about high throughput and low latency, while both are not there.
one cannot bridge or act in reality for 32 * 400 ms
https://solana.stackexchange.com/questions/7945/single-slot-...
To date no transaction has been rolled back on Solana once two thirds of the validator set have voted on it.
tl;dr Users often submit transactions that attempt to arbitrage price discrepancies on-chain. But someone else fills the arb before them and so their transaction while successfully received will result in nothing happening. The transaction didn't fail though. Solana, your "decentralized server", successfully received the request.
The result is that everyone relies on the Solana Foundation to store the full archive, and pretty soon even non-expired history will only be possible for large companies to store.
for validators it may be last 300 blocks. for bridges 30k blocks. for rollups 300k.
assuming it is possible to start validator from any block(trusting random subset of validators to bootstrap) and account state is way smaller its ok.
also, as i see 100tb costs 3k usd. so some archieval nodes can be just payed some to handle that.
And that number is going to grow exponentially, to the point where the network is destined to have a centralized, and thus fragile, structure.
I agree that most nodes do not need the full history, but even everyday operational costs are enormous: dozens of gigabytes a day just to keep up?
so economics of storage i do not know, but that leads back to artificially cheap transactions. as soon as tx will be real price, they will pay storage.