I agree changing the protocol (I'll assume you are only talking about consensus changes through soft/hard forks) is very difficult, CTV has been proposed for several years now, and I am witnessing it first hand myself. As for it being very dangerous, this totally depends on what the change is. A new OP code is one of the smallest changes possible, far less complexity than segwit or taproot were. CTV is also the most restricted op code that is currently proposed (you will struggle to find a serious dev that disagrees), it is non recursive (best to look this up if you don't know this term), and has a relatively small amount of code. The code has been reviewed for quite a few years and heavily tested, with a 5+ BTC bounty available for anyone who can create a damaging tool/bug/etc using CTV on bitcoins testnets. This has also been available for several years with nothing found so far. CTV could certainly be used for bad things where you could lose your money, but you would have to volunteer to send your bitcoin to an address that uses CTV (so it wouldn't affect any other users). This is the same as now, you could send your bitcoin to an address you don't control or you could use a time lock to lock your bitcoin for thousands of years, etc. There will be new code to maintain of course, but CTV will also improve efficiency of lots of current code/use cases making some things simpler. As for the rules not being meant to change, firstly Satoshi would have disagreed, you can see this in his writing. And also bitcoins history disagrees, as there have been 8 consensus rule changes since March 2012 (see the attached image). Also non consensus code changes very regularly, check out the bitcoin GitHub. There are very good arguments that covenants like CTV can improve the economic incentives of the system. Such as improving the fungibility of coins with coin pools (like using coinjoins but inside a single UTXO), and increasing the economic density of a UTXO so monetary transactions can better compete with things like ordinals or inscriptions (which some call spam).

Ok now for the second point. I'm not very convinced by this personally, and I believe I went into this a bit in the post. If lots of coins end up TRAPPED (due to fees to exit being too high or the custodian refusing/legally not being able to let you exit) you no longer have the same free market we have right now. We could not force bank runs on them to keep them in line. This would be a similar system to what we have now with fiat, there is some competition amongst banks, but it's far from great (as I'm sure you know). We would probably be back to having to protest or use the legal system. The open ledger could provide a slight improvement though on seeing their reserves (assuming they share their addresses). But we could never truly know their liabilities, so they could create fake IOU bitcoins and go back to fractional reserves. As for their authority on the bitcoin protocol, I feel I covered that well in the post, but I'll touch on it again. If they get enough power through their nodes and they collaborate with other custodians (the government they reside under could force them to do this, the custodian themselves doesn't need to be a bad actor).