They could block Snowflakes with IPs from networks in unsafe countries, but that is trivially bypassed by the attacker just buying VPSs (or botnet nodes) in a freer country.

Skimming the Technical Overview[0], I don't see anything about mitigating the risks you mention.

The purpose of Snowflake seems to be to circumvent blocking of Tor, not to prevent detection of using Tor. It takes advantage of "Domain Fronting" and WebRTC to accomplish this.

[0] https://gitlab.torproject.org/tpo/anti-censorship/pluggable-...

> that is trivially bypassed by the attacker just buying VPSs (or botnet nodes) in a freer country

A.K.A. "living off the economic land"

"Just" don't connect from an IP that can be tied back to you, use black market sim in a separate phone, connect from places you don't go, turn it off when not in use... It gets expensive fast...

> use black market sim in a separate phone

In most countries this takes you from “may have committed a crime” to “have actually committed a crime”

We're talking about countries where using vpn is already a crime so no problem

In most places where Snowflake is useful, connecting to Tor is either legal or the laws against it aren't enforced. It's usually the creators/contributors of anti-censorship tools that face repercussions. That said, Tor Project pretty consistently emphasizes that all plugable transports are for AC purposes, not steganographic purposes, and while they're difficult to block, they will not stop the network operator from being able to tell you're connecting to Tor, and that it ultimately falls on the user to decide whether that's acceptable.