Leaving out the specifics of how much you would take, if you want to go down this route then you would need to create some kind of automatic mechanism to take these coins from people, and not rely on them spending them as enforced fees, as there would be an incentive for individuals to let these coins be excluded from the circulation as it would make their other coins more scarce and therefore more valuable.
Yes. And this should show clearly, what tail supply is really about: it is about taking single satoshis from all accounts, no matter what, no matter if coins are burned or not, no matter if someone wants that or not. So, it is all about creating an invisible tax on all addresses. If some coins are fully burned, for example by using OP_RETURN, then they cannot be moved. But tail supply is about taking single satoshis from those addresses as well.
So, you know what is needed: zero satoshis. Then, it is possible to create some additional outputs, send zero satoshis there, and use "<anyStandardScript> <newAmount> OP_DROP" as an output script (or this "<newAmount> OP_DROP" could also be placed inside witness script, or as an input, many things are possible). It could be handled in the same way as Segwit vs NonSegwit: if it was possible to create a situation, where old nodes cannot see new signatures, then it is also possible to create a situation, where old nodes will not see new amounts (there could be many reasons, for example if hiding amounts will ever be introduced, then it is reasonable to put zero for backward compatibility, but the same solution can be used to introduce any coins to the system, because the size of the UTXO set is not limited). And then, it is all about human factor: if those zero satoshis will be really used to move real values, then they could be traded, bought, sold, and used in real life. If it is possible to create NFTs out of thin air and sell them for millions, then why producing coins out of thin air and selling them for real goods and services wouldn't work as well?
what are the main arguments against using merged mining to create security once the block subsidy is insufficient?
The main arguments are that sidechains can be unsafe. So, that means they will be created anyway, there are many options, how exactly it would happen:
1) it can be deployed on altcoins
2) it can be deployed on second layers like the Lightning Network
3) it can happen on centralized websites, like exchanges or casinos
4) it can turn out that homomorphic encryption is sufficient to deploy sidechains on the main network
5) other features can enable sidechains by mistake, just because developers can be unaware that some feature enable more things than intended
So, I think it will happen anyway, that way or another. Definitely, homomorphic encryption is a powerful tool, because it can enable new features in a permissionless way.
Is there any ongoing discussion or work on blind merged mining protocols (such as BIP 301)?
Yes, you can talk directly with Paul Sztorc and other people on Telegram.