It is a lazy way to cater to the lowest common denominator of users who will eventually fall for a phishing scam or install some keylogger and have their password end up in a dump.
I've come to that conclusion, too. I moaned and groaned when Microsoft bought GitHub, but did nothing about it. Now I've begun the migration process.
I'd like to spell out some big-picture thoughts on 2FA. Security and convenience are antithetical to each other. Sometimes better solutions come along where you can have a little bit of both. But ultimately, they're trade-offs.
Do you value security, or do you value convenience? And in what proportion?
Furthermore, increasing security also increases fragility. There's more to go wrong. Posters here have been talking about losing their devices, and so forth. These are legitimate concerns. These points have rebuttals, such as recovery procedures, and so forth. OK, and you're sure grandma is up to the task, is she? All these convenience tools: they're easy to use until they aren't.
So not only do you have to think about security vs convenience, you have to consider downside risks on both ends of the spectrum. It's not as easy as saying "moar security" and reading press reports about the advisability of such systems.
Or to look at this from the threat analysis side, there are many different threat models and not everyone has the same ones.
This is why it is so wrong for these cloud services to consider only one single threat model and impose it on everyone.
Denial of service is also a consideration in threat modeling and 2FA does increase the risk of loss of access. Whether that's worth it depends entirely on other factors which will vary for each individual and group. It's never a clear cut answer.
Personally I have TOTP enabled on my github account (not on a phone, that's too fragile) but that's because it makes sense for my use case.
There are other provider accounts where I have no desire to ever enable 2FA because I value access from anywhere with only my memory far more.
Cloud providers should not be making threat modeling decisions on behalf of users, each user needs to make their own.
Another reason to self host most things, you can implement your own best policy.
IMHO there’s no good answers on global scale. Identity should be handled on local level. Government already has process for issuing me new tokens even if I would loose all my existing ways of proving my identity. Local organizations know how to verify my identity using those tokens. I already put lots of trust on my own bank on this, so maybe they could also manage my digital identity.
Another solution is for users to upload identification documents when they sign up for the service. To recover your account, the service asks you to provide the documents again. This way it doesn't matter what the locality is, and you can provide literally anything (a picture of a duck!). Some providers have asked me to send a copy of my Driver's License before, but I don't think I had provided it to them before recovery, so that wasn't great.
A couple startups are beginning to build "identity" products that I imagine will cover a lot of this space. I expect soon there will be one company that everyone uses for identity, and then rather than be at the mercy of GitHub, we'll be at the mercy of Sauron's Eye.
GitHub is a remote. This "mirror" is kept on your disk.
I think we all need to consider the possibility of moving off of GitHub, or at least keeping a mirror of everything on another provider, and making sure any long-lived services that pull from GitHub know the other provider to use. You don't want an account lockout to mean you've lost all your work.