To be honest, CRDT sounds like a fancy word for blockchain, and you don't have to be an OG to figure that out.
I looked it up and found this article:
For the application level, which represents transaction processing and validation logic, we defined the following set of buckets to hold the state:
Epochs — a KV subspace (bucket) that holds epoch snapshots. An epoch snapshot is similar to a block in the classic blockchain, but within our network transactions are grouped in epochs. More on that below. Each epoch has a number as a key that points to a list of transactions.
Transactions — a bucket holding set of confirmed transactions. Consensus participants decide whether or not to include a transaction into that set. If the transaction is confirmed, its CID and validator signatures are added to the set, and the body (data, metadata, result, etc) will be replicated through IPFS using DAGSyncer.
So from a birds eye it's just putting transactions on the chain after every epoch, and not every time a block is generated. An epoch is 2 weeks in bitcoin so that would make transactions prohibitively slow to confirm. Of course I could be wrong about all this because I just glanced through the article.
(DLT just stands for distributed ledger technology.)