This degree of skepticism can be counter productive.
If everyone in a democracy agrees that "governments would never allow" large companies to implement end-to-end encryption, then, as a result, governments never will.
Also, some large companies do offer it, notably Meta and Apple. Apple offers a backdoor in the form of unencrypted iCloud backups. And Meta may also offer a backdoor to the US government, but they also might not. And there are at least a few large governments they don't appear to offer a backdoor to, as evidenced by Meta executives getting arrested for contempt of court in e.g. Brazil.
Twitter encrypting DMs is definitely an improvement, just as Twitter adding SSL was an improvement.
Security and privacy ultimately depends on threat modeling. Any given improvement can make the difference between the suitability or unsuitability of a tool for a given threat model.
And since peoples' threat models are extremely diverse and always changing, adding e2ee to a service of this scale can have a big impact on the security, privacy, and safety of many Internet users.
I am both an idealist and a realist. I'm very much aware of how much surveillance governments run on their own citizens. There's no place in the world that is exempt from this. So you could either move to Mars and hope the Elon doesn't do things like everyone else, or you can try to effect change here, but it's an uphill battle. All governments are drunk with the power they have, and they are not going to give it up willingly.
And yet if you look at the state of the art in surveillance for most governments in the world (buying tools from NSO Group) it seems that many governments have given up on mass surveillance of messages in transit, and instead are focusing their resources on targeted surveillance through device hacking.
So some power has clearly been given up. Not willingly. But still it has.
And in the Snowden leaks you see that even the NSA had problems with tools like Tor and Tails.
Granted that was a decade ago, but there's no reason to believe that work by privacy activists to promote adoption of practices like SSL and START-TLS, or tools like Signal, has not had some impact on the cost and availability of mass surveillance to governments.
I'm not sure where you get your information, but the only reason a government would abandon "in-transit" surveillance is if they have full coverage from another source.
If everyone in a democracy agrees that "governments would never allow" large companies to implement end-to-end encryption, then, as a result, governments never will.
Also, some large companies do offer it, notably Meta and Apple. Apple offers a backdoor in the form of unencrypted iCloud backups. And Meta may also offer a backdoor to the US government, but they also might not. And there are at least a few large governments they don't appear to offer a backdoor to, as evidenced by Meta executives getting arrested for contempt of court in e.g. Brazil.
Twitter encrypting DMs is definitely an improvement, just as Twitter adding SSL was an improvement.
Security and privacy ultimately depends on threat modeling. Any given improvement can make the difference between the suitability or unsuitability of a tool for a given threat model.
And since peoples' threat models are extremely diverse and always changing, adding e2ee to a service of this scale can have a big impact on the security, privacy, and safety of many Internet users.