This argument can be very easily turned around. For people like that, isn't it better to know they have been exposed? Monero is perfectly happy for them to be tracked by Ciphertrace while reassuring them that all is fine.
You've convinced
me that IP addresses for a significant chunk of Monero TXs have been compromised but right now that data is silo'd in ciphertrace/chainalysis/Fourteen Eyes servers (and yours). I realize you were able to code this up mostly on your own but I think you'd be surprised at how far behind the intelligence agencies of most countries are so some Turkish or Azeri or Libyan dissidents or something could easily be affected by this (that being said, it is possible ciphertrace is selling it to Turkey through Cellebrite).
I think most people aren't knowledgeable enough to authenticate your data and are assuming it's FUD. If you want to quickly get the word out, an easy way might be to contact a reputable journalist and demonstrate it to them by revealing their own IP address (I'm happy to help you with this).
There is no such thing as a public mempool on a p2p network level. What you are seeing on the website is a snapshot of one node's mempool. The way dandelion++ works is that the transaction will be put into the original node's mempool, broadcast in stem phase, and if the node doesn't see it again, it will be broadcast in fluff phase. The "not seeing it again" part is the cause of the delay.
Interesting thanks for the clarification. I'm not particularly familiar with dandelion++. Does the original node broadcast the transaction to multiple nodes in the stem phase or just to one other node? When does the fluff phase occur?