Handling the non-1:1 case when the 1:1 case is solved is trivial, more challenging is figuring out actual applications where users would opt-into it. There are silly ones, e.g. "ponzi chain"— I guess one metric for an approach being general and supporting of freedom is that it can accommodate things that sound really ill advised too.
.. but I've not really found many idea in the non-1:1 space that I'd realistically expect many to play along with.
One argument might be that it's potentially desirable to have a system where lost coins eventually go to support the system somehow... but that very easily runs into issues with bad incentives as miners can censor transactions to make coins seem lost when they're not.
but can merge-mining be used in the reverse way
Depending on what validation rules you choose you can get whatever security "umbrella" you desire. E.g. the sidechain can have consensus rule that their work must also commit to the best valid Bitcoin chain (effectively importing bitcoin validity into the secondary system). This degrades the tidy isolation that you normally get with no linkage, but it gives full security, complete with incentives (because if you misbehave in the Bitcoin blocks you also lose your blocks in the other system). This is good news, but potentially not a cure all— in the you may still be failing to meet your decentalization objectives even though the hashrate security (ignoring centralization degree) is good (e.g. because the scaling requirements and/or goals of the more profitable secondary system are less decentralization compatible).