CoinJoin works as a method of both improving fungibility and mixing with
coinbase transactions.
My understanding is that the two situations are quite different.
Unlike mixing to coin-split, CoinJoin doesn't create a high demand exclusively for coinbase transactions.
However, of the proposed methods, coin-mixing seems the better option, because it might be reasonably easy (I don't know) for exchanges to obtain 148 coinbase coins, and mix their coins with them, extending the coin-splitting capability beyond just miner coins and then using that to split incoming coins.
That seems like the most reasonable approach I've heard so far. Whether exchanges would be willing to do that is a separate question.
When it's confirmed on one chain, but not on the other, you
can then "double-spend" on the lower hashrate chain with a higher fee,
to end up with different coins on both chains.
This method is time consuming and not guaranteed to work. CPFP can be used by an attacker to get your original txn into the 148 chain.
(also, no double-n in untenable)
Why thank you aj, you're so good at spelling. :-)
Cheers,
Greg