

Pretty easy honestly.
You do something like remove section 230 (or whatever the EU equivalent is) that provides safe harbor from liability for transit providers like ISPs and content providers like websites that host user submitted content. You condition any safe harbor on the services in question being able to turn over and ID exactly who the offending person was without fail and tie any and every packet to a real world person. You make explicit that not being able to scrutinize content (because of encryption) is not an excuse. Thus someone pirating or sending CSAM over your network via a VPN makes you liable for not stopping them.
As a result this forces ISPs to block all encrypted traffic detected via deep packet inspection. Only traffic encrypted with public key infrastructure that has government issued keys that allow snooping on it is allowed.
Tada. There’s no way around this that doesn’t involve painstaking steganography which can possibly be nailed by AI anyways. Things like embedding a secret message in pictures you send with some pixels shifted to hide the data and your friend having a program and key that can decode it. Or things like taking all the capitalized letters and applying rot13 or something to them with some sort of algorithm but then you need to find a way to make the message intelligible on the surface as if you’re sending constant unintelligible messages you might get flagged and blocked or visited by the police (or the police get a warrant and have your mobile company deploy malware onto your devices and spy on you as a threat because of that).
The only other alternative is using alternative infrastructure. HAM radio type network transmission via a series of hops with similar activists but this wouldn’t be practical for most given the expense and the bandwidth would be awful. Also probably illegal and if they wanted to it would be trivially easy to identify and arrest those running these nodes and relays due to triangulation.
Turns out the whole liberal west with freedom of thought and speech was in fact a lie. Kept around to use as a stick to whack at the USSR with but now dropped at the first signs of serious popular discontent and trouble in favor of total control. Supposed values quickly dropped with no more excuse than “Russians” or “think of the children” or the usual criminals and terrorists.
They can’t stop a really determined actor from engaging in encrypted messaging but they can stop 98% of the population and that’s more than enough to control thought and action of the population.
No. It’s fine.
Tor uses its own DNS system to my recollection. It’s true there is DNS as part of fingerprinting and DNS leaks are a concern for VPNs (see for example https://www.dnsleaktest.com/) but Tor is not vulnerable to this and it’s more a problem of you’re using a VPN to appear to be in NYC but your DNS shows Phoenix so that’s a big discrepancy that raises the uniqueness of your fingerprint on a VPN and even lets threat actors guesstimate where you actually are. As I said though this is not an issue on Tor.
So understand that the DNS from Mullvad will only affect other programs not Tor. It will prevent say your ISP’s DNS from seeing your video games calling their domains that way. Your ISP can still see you’re connecting to infrastructure for as an example Genshin Impact when you launch the game because they can see where your traffic is flowing and the IP addresses as well as traffic patterns, ports, etc. It somewhat limits the data and visibility they get but there is something called SNI snooping as well as of course the fact they know the IP addresses where your connections go. So it’s perhaps better than nothing but understand the limits of it as they still have a lot of visibility though they shouldn’t be able to see your web searches regardless just that you’re accessing google or bing or duckduckgo as those sites use HTTPS.