I think you are confused as to the purpose of the filters. You said yourself that knots runners didn’t have this transaction in their mempool, so sounds like the filter worked. You just got a node that had a higher limit filter or no filter to put it in a mempool.
This 2679 byte op_return containing a png file was confirmed on chain. I broadcast it using Core version 29.1 to the regular Bitcoin relay network (no out of band service needed). Core 30 makes no difference in the ability of a "spammer" to confirm transactions with large op_returns.
Here is the TXID: 0d5f273d09c1a4665634fd25d5a17b879d8843de5edc40b8b7d8671500dd16b6
It took a little while to be mined, I must have gained a new peer that would relay it to a miner. I broadcast it block height 916259 and it was confirmed at 619343.
I paid less than 1 Sat/vB. Knots users didn't have this transaction in their mempool but now they store it on their node. Filters won't keep these transactions off your node. Filters do not work.
You can view it for yourself with this command (Thank you nostr:nprofile1qqsdnpcgf3yrjz3fpawj5drq8tny74gn0kd54l7wmrqw4cpsav3z5fgpzemhxue69uhk2er9dchxummnw3ezumrpdejz7qg4waehxw309aex2mrp0yhxgctdw4eju6t09uq3jamnwvaz7tmjv4kxz7fwwdhx7un59eek7cmfv9kz74rw7n6):
bitcoin-cli getrawtransaction 0d5f273d09c1a4665634fd25d5a17b879d8843de5edc40b8b7d8671500dd16b6 true | jq -r '.vout[0].scriptPubKey.asm' | cut -d ' ' -f2 | xxd -r -ps | base64 -d > filters.png
Here are my node settings that allowed me to send this transaction and get it relayed (and mined):
minrelaytxfee=0.00000100
mempoolminfee=0.00000100
incrementalrelayfee=0.00000100
datacarriersize=100000
If you are running Knots because you believe it will prevent "spam", you are being fooled by Mechanic. The only way to keep this type of transaction off your node it to fork #Bitcoin consensus rules.
Discussion
And you potentially downloaded and validated the transaction twice if it happened to be relayed to you. Sounds like an ineffective filter.
I guess Luke could build in the option to only peer with other Knots nodes but some Knots nodes would still have to peer with Core nodes.
Can you explain that to me? Why would it be twice? Could it be more than twice? Or not at all?
But do you agree it wouldn’t be in my mempool?
If my node happened to relay a transaction with a large OP_RETURN to your node, you would download the transaction from me, validate the transaction, and then determine that it violated your mempool policy and drop the transaction. Your node wouldn't relay the transaction to any other nodes. Then if the transaction were to be mined into a block (because it is consensus valid) your node would download and validate the transaction a second time.
If you happened to be peered with several #Bitcoin Core 30 nodes it is possible that multiple nodes would relay the transaction to you. I'm not sure if the Bitcoin software has an efficient mechanism in place to drop a transaction it has previously seen and determined to violate the mempool policy or if it will repeat step 1 again of downloading, validating, and dropping. If you were peers with mostly Knots nodes you would likely not see the transaction at all until it ended up in a block.
To answer your main question, my transaction with a large OP_RETURN would never be in your mempool.