I confess that I'm still confused.
The non-RBF transaction is invalidated with double-spending its unconfirmed parent (which is RBF-enabled).
In this case, shouldn't the CoinJoin/Lightning software warn for unconfirmed parent(s)? I wouldn't accept a non-RBF transaction which has an unconfirmed RBF-enabled parent as "settled".
With invalidating the parent, the child (which was spending the same UTXO as the coinjoin transaction) is removed from the mempool and the coinjoin transaction can be propagated and confirmed.
You mean
can't be propagated and confirmed?
I make Transaction A, which spends UTXO 1 and creates UTXO 2, and is opted-in to RBF. I then use UTXO 3 to join a multi-party transaction. At the same time, I make Transaction B, which spends the unconfirmed UTXO 2 and UTXO 3 and is opted-out of RBF, and broadcast Transaction 2 to the rest of the network.
You mean Transaction
B? You haven't mentioned of any Transaction 2.
Miners are generally very well connected to a myriad of nodes. However, the main concern should be whether the miners are willing to change their policy and start accepting these kind of transactions.
Isn't compulsory RBF benefiting miners? Why wouldn't they want to change their policy accordingly?