Fair enough, but stamps uses counterparty, and the counterparty FAQ says this about OP_RETURN and prunability: https://docs.counterparty.io/docs/basics/faq/#what-happens-if-and-when-op_return-data-is-auto-pruned

> Counterparty only needs some Bitcoin full nodes somewhere to have an unpruned copy of the blockchain. As every Counterparty full node is also a Bitcoin full node, this is easily done.

stamps references the counterparty behavior as the reason why stamps uses bare multisig and says

https://github.com/mikeinspace/stamps/blob/main/BitcoinStamps.md

> The length of the string means that Counterparty defaults to bare multisig, thereby chunking the data into outputs rather than using *the limited* (and prunable) OP_RETURN.

Since counterparty refutes/denies the pruning argument, it's IMHO safe to conclude that stamps is actually using bare multisig because of the datacarrier limit anyway.