If someone publishes code for themselves, and at no time asks anyone to take it as a dependency, then at a later date they change that code in a way that breaks other people's use of it, do GH then take over the account?
If the intent of the push was to damage downstream users of the software then it is malicious towards them.
This is like arguing about whether the james webb telescope really is in space since we don't have a precise consensus about what altitude is considered the frontier with space.
To take a trickier example, say a GH user has a lib, then decides to re-architect it, breaks the API and for their own purposes pushes it to an existing version, breaking all other use of it. Now that's a nasty thing to do, but is it malice?
Another, real-world, example is I know of a user who publishes "honey PoCs" for security issues, where the repo. appears to be a exploit code but actually isn't. He's been accused of malice in doing this, but his intent is research for a talk on how people use code blindly without testing.
Is that malice, should GH take his account down?
By stepping into this area GH are going to have to find answers to this and also the problem of who maintains the repos of accounts they nuke?
I think they already have, those 2 examples you mentioned already happened and were dealt with.
I think intent is important to take into consideration, since after all that is the definition of malicious: intent to cause harm.
Your first example clearly has no intent to cause harm. That case probably happened thousands of time already since not everyone is willing/able to follow semver cleanly and strictly. Never heard about GH taking any measure against that. And I would definitely not expect them to as a user/maintainer.
For the second case, I think GH policy is that you can host that kind of PoCs, but the repo has to be clearly documented as doing such (e.g. you can't just add some vuln into some unrelated code "for research"), and the vulnerability cannot be an active one: "We understand that the publication and distribution of proof of concept exploit code has educational and research value to the security community, and our goal is to balance that benefit with keeping the broader ecosystem safe. In accordance with our Acceptable Use Policies, GitHub disabled the gist following reports that it contains proof of concept code for a recently disclosed vulnerability that is being actively exploited." - GitHub." [1]
Back to Marak's case, my opinion is that GH did the right thing: If he just had his code in a repo, with no semver, no other contributors/maintainers, and such, and decides to nuke it, then I hope GH would not have done anything.
But when you are using all the tools and trust of open source: Other people contributing to your repo, other people being active maintainers/admins and spending times out of their days to fix bugs on it, when you leverage NPM to make it easier for you to distribute your package to others widly etc, you give up the privilege being able to act unilaterally like an asshole without consequences.
[1]: https://www.bleepingcomputer.com/news/security/githubs-new-p...
For example in that first case, say all they had was a wave of people saying "x broke my application" it would look a lot like the case in the article, and they'd have to dig in to find out it was just a bad API change without semver being followed.
Also requires Github to have a staffed department to deal with this, now they've established themselves as the arbiters.
For me, there's a split between a repository (NPM) and a hosting company (Github). For this case I'd have forked the repo, rolled back the malicious change in the fork, and hooked the fork up to NPM, and leave the original GH account alone. That solves the problem of the breakage, without getting in to banning whole GH accounts.
I'm (honestly) trying to understand why GitHub can't just go ahead and remove any account as they see fit.
Edit: Am I being downvoted for asking a question?