Since transactions with larger OP_RETURN outputs or multiple OP_RETURN outputs are consensus valid, they will appear in blocks as long as they find their way to mining pools and mining pools choose to include them. Any nodes that don’t allow such transactions in their mempool will download them twice: once when the transaction is first offered to them by a peer, where they download it, evaluate it, and drop it due to their mempool policy, and then a second time when they receive the block announcement that includes the transaction.

Whenever nodes are missing transactions from a compact block announcement, they have to request the missing transactions which increases the latency until they can relay the block. The relay is delayed more when large transactions are missing. Nodes that run a mempool policy that is stricter than what regularly appears in blocks therefore use more bandwidth and relay blocks more slowly.

Slower block propagation benefits larger mining pools by increasing the number of stale blocks and delaying other miners in switching to the new chaintip. Bigger miners disproportionally win over competing blocks and the block author doesn’t suffer from the propagation delay when a block is found.