Software doesn't violate licenses, people do. Anyone who violates the license of a work loses their license to the work. A more pointed question then is about who violates/loses the license: an individual coder? A corporation/foundation?

I think the gist was: if you lose your license to distribute Busybox X, does that impact your rights with respect to Busybox Y?

My take on this (but IANAL) is that you do not really lose the rights to distribute a specific version, but the rights to distribute specific code (or compiled versions of it), and that would carry through to all code in version Y that was already in version X, but not any newer code that is unique to version Y. That won't be very useful, though.

[Edit] Just read the part of the article that deals with this question. Personally, I hold to my take on this matter -- that the scope is not a particular version of the distribution containing the code, but to all the bits of code of which the license was violated -- even if you also have access to the code in another way. My reasoning for this is that this could otherwise open up a pretty simple loop-hole: you'd only have to get someone else, whose license is not yet revoked, to fork the project, and release a new 'version' of the program, to get your rights to the code back. That can not have been the intention of that clause in the GPL.

Ah I thought you were naming BusyBox as an infringing work, not an infringed. Anyway, it does seem, as mentioned in the article, that if the copyright holder rereleases a similar work under a similar license, a past infringer gets a fresh shot, unless the author specifically publishes their revision history as a single work.

The point of the GPL really is about sharing, as the article suggests, not about punitive damages or post-violation injuctions. GPL-using authors do not want to prohibit infringers from future sharing, they want derived works published freely.

Much of the point of GPL and Free Software is about freedom for users, not making war against others. A past infringer gaining access to to code is not a loophole, it is part of the goal of the Free Software movement, which is unrestricted access for anyone to exploit software privately or to share it publicly (but not to allow public exploitation of non-shared code).

I think the position that it's about specific code is not defendable. What if I delete the file and retype it as a coder (see Theseus)? Is this the same file that was in a former release? If I automatically code format it, so no character is at the same position as it was before? Is it the same file? The only thing that makes sense is considering releases as the particular thing that is licensed (with each source file being a considered a part of the release, not a part in the history of the file).