I mostly agree with you, but 23andme could have prevented this by requiring 2FA for all accounts, or at least for accounts with an email in the HaveIBeenPwned database.
Credential stuffing is most preventable by the user (who can simply not reuse passwords), but platforms have a responsibility as well. They can at least mitigate it through rate limiting, and mostly stop it with 2FA requirements.
If an attacker is able to exfiltrate millions of records from a platform with credential stuffing, that means they tried to login to multiple millions of accounts. It shouldn't be difficult for a service to detect and stop such a sustained level of load on its login infrastructure. You can't get millions of proxies.
Credential stuffing is most preventable by the user (who can simply not reuse passwords), but platforms have a responsibility as well. They can at least mitigate it through rate limiting, and mostly stop it with 2FA requirements.
If an attacker is able to exfiltrate millions of records from a platform with credential stuffing, that means they tried to login to multiple millions of accounts. It shouldn't be difficult for a service to detect and stop such a sustained level of load on its login infrastructure. You can't get millions of proxies.