Unpopular opinion
With current bitcoin opcodes many layer 2 solutions can be built as long as the validation is on the client side.
Face it, it is what it is.
Discussion
I would like to see certain things happen in Bitcoin, but decentralization is what it is, for better or worse.
Build them on Liquid.
You don't like Liquid? then come up with something new.
You think it's bad? again I say, it is what it is.
Apply stoicism to Bitcoin.
Yes, but this doesn't *necessarily* imply custodial solutions for higher layers. The question is whether there are crypto techniques that can "latch" verification of something complex, to something that exists within Script. And yes, there are; it's just no one's found really good ways to make use of them, yet. You had ZKCP from Maxwell back in the day, which is pretty much "the" canonical way to do it (transaction unlock contingent on revelation of arbitrary witness) then sort of "indirect" verif of cryptographic claim in e.g. adaptors (recently i posted about how you could even do a kind of DLEQ verification onchain, on the mailing list), then there is the radically different game theoretic approach ("optimistic" i.e. punishment based) with bitvm, which I can see making *some* headway but not being "the" solution. I do agree with your controversial take though; it may be that we *need* to avoid the requirement for changing op codes. I have always intuitively felt that to be true; 2017 was somehow a kind of cut off.
The chain can only protect itself. Everything else requires separate enforcement, and once established this new authority will eventually become its own source of truth.