Since, realistically, neither Linux nor GNU will budge, maybe the practical solution would be to look into integrating with the equivalent library in the LLVM ecosystem (assuming it exists).
The GPL 3 is a pretty benign improvement on the GPL 2 (though I know Linus objects to the Tivo clause). I don't think anyone who was happy with their software being used under the terms of GPL 2 would be unhappy with it being used under the terms of the GPL 3 additionally, rather than exclusively.
But let's imagine the next version of GPL 4 was "Additionally the authors may at their discretion use it under the terms of the CC-BY-SA license". Nothing wrong with it as a license, but that's a bigger leap in terms of license changes.
It's not hypothetical that they could exercise that kind of power - they _did_ with the GFDL with the "wikipedia can relicense as CC-BY-SA" clause in GFDL 1.3.
I think most people would agree that that was a responsible use of that power. but I can understand those that don't want to extend the trust that all future uses will be responsible.
“You can do A, B, and C with the software I wrote, or whatever X may at some future date decide you can do with it” isn’t something everybody is happy with. It certainly requires some long-term trust in what X will or will not do.
If the FSF goes rogue and changes the GPL to be incredibly restrictive (i.e., allowing proprietary redistribution, and I realise this can be considered permissive..) it might be possible to get it to be ruled invalid defaulting to the more permissive licence, especially if you have deep pockets, or if the FSF change the licence to be ridiculously permissive like 0BSD then it's not going to be legal in countries like Germany, either way any major change is likely to result in an international enforcement nightmare.
The scenario where FSF goes "rogue" would be if gplv4 became more permissive, e.g. if it stopped being a copyleft licence.
It says "the recipient can use this under GPL v2 or any later version as published by the FSF", which means if they recipient is happy with the rights they got under GPL v2 they can keep using it under that version for as long as they like.
The only person agreeing to future license changes is the publisher of the code, and they're the ones that chose to publish it under "GPLv2+" or whatever.
It doesn't make sense to assign blame on those who bug-fixed the backdoor in the license that clearly would allow a third party to change licensing terms. Even(!) if that third party was rms.
>The blame is pretty clearly on the people who editted the license text to be GPL2 only.
They edited the licence, yes, but the FSF explicitly wants you to do this to make your intention clear(1). When you licence software under the GPLv2 (or 3, etc) you have a choice of 'GPLv2 only' or 'GPLv2, or any later version', however since the licence text only states 'Version 2' with the old short labels being just 'GPL-2.0' there's some ambiguity on whether you mean GPL-2.0-only or GPL-2.0-or-later.
The default assumption should always be v2-only, however as (at the time) the FSF were still recommending the short label of GPL-2.0 and the issue of using v2-only or v2-or-later wasn't really an issue you had a lot of v2 licenced software and patches using the default unedited licence with the FSF short label of GPL-2.0 and this is purely the fault of the FSF. It wasn't until the GPLv3 came around which some people didn't like (notably the Linux kernel, which is probably why perf is v2-only) that you got people editing their licences to make the intention clear, although for many projects they had no choice in the matter as changing to v2-or-later would require permission from every copyright holder that had contributed code to that project, again this is partially the fault of the FSF for not having enough foresight or making the choice of -only or -or-later more explicit and clear.
P.S. if you already know the history and context here this post probably seems a little patronising and I apologise for that.
(1) https://www.gnu.org/licenses/identify-licenses-clearly.html